-/*-
- * BSD LICENSE
- *
- * Copyright 2015 6WIND S.A.
- * Copyright 2015 Mellanox.
- *
- * Redistribution and use in source and binary forms, with or without
- * modification, are permitted provided that the following conditions
- * are met:
- *
- * * Redistributions of source code must retain the above copyright
- * notice, this list of conditions and the following disclaimer.
- * * Redistributions in binary form must reproduce the above copyright
- * notice, this list of conditions and the following disclaimer in
- * the documentation and/or other materials provided with the
- * distribution.
- * * Neither the name of 6WIND S.A. nor the names of its
- * contributors may be used to endorse or promote products derived
- * from this software without specific prior written permission.
- *
- * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright 2015 6WIND S.A.
+ * Copyright 2015 Mellanox Technologies, Ltd
*/
#include <stddef.h>
-#include <assert.h>
#include <errno.h>
#include <string.h>
#include <stdint.h>
#include <fcntl.h>
#include <sys/queue.h>
-/* Verbs header. */
-/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
-#ifdef PEDANTIC
-#pragma GCC diagnostic ignored "-Wpedantic"
-#endif
-#include <infiniband/verbs.h>
-#include <infiniband/mlx5dv.h>
-#ifdef PEDANTIC
-#pragma GCC diagnostic error "-Wpedantic"
-#endif
-
#include <rte_mbuf.h>
#include <rte_malloc.h>
-#include <rte_ethdev.h>
+#include <ethdev_driver.h>
#include <rte_common.h>
#include <rte_interrupts.h>
#include <rte_debug.h>
#include <rte_io.h>
+#include <rte_eal_paging.h>
+
+#include <mlx5_glue.h>
+#include <mlx5_malloc.h>
+#include "mlx5_defs.h"
#include "mlx5.h"
#include "mlx5_rxtx.h"
#include "mlx5_utils.h"
#include "mlx5_autoconf.h"
-#include "mlx5_defs.h"
-
-/* Initialization data for hash RX queues. */
-const struct hash_rxq_init hash_rxq_init[] = {
- [HASH_RXQ_TCPV4] = {
- .hash_fields = (IBV_RX_HASH_SRC_IPV4 |
- IBV_RX_HASH_DST_IPV4 |
- IBV_RX_HASH_SRC_PORT_TCP |
- IBV_RX_HASH_DST_PORT_TCP),
- .dpdk_rss_hf = ETH_RSS_NONFRAG_IPV4_TCP,
- .flow_priority = 0,
- .flow_spec.tcp_udp = {
- .type = IBV_FLOW_SPEC_TCP,
- .size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
- },
- .underlayer = &hash_rxq_init[HASH_RXQ_IPV4],
- },
- [HASH_RXQ_UDPV4] = {
- .hash_fields = (IBV_RX_HASH_SRC_IPV4 |
- IBV_RX_HASH_DST_IPV4 |
- IBV_RX_HASH_SRC_PORT_UDP |
- IBV_RX_HASH_DST_PORT_UDP),
- .dpdk_rss_hf = ETH_RSS_NONFRAG_IPV4_UDP,
- .flow_priority = 0,
- .flow_spec.tcp_udp = {
- .type = IBV_FLOW_SPEC_UDP,
- .size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
- },
- .underlayer = &hash_rxq_init[HASH_RXQ_IPV4],
- },
- [HASH_RXQ_IPV4] = {
- .hash_fields = (IBV_RX_HASH_SRC_IPV4 |
- IBV_RX_HASH_DST_IPV4),
- .dpdk_rss_hf = (ETH_RSS_IPV4 |
- ETH_RSS_FRAG_IPV4),
- .flow_priority = 1,
- .flow_spec.ipv4 = {
- .type = IBV_FLOW_SPEC_IPV4,
- .size = sizeof(hash_rxq_init[0].flow_spec.ipv4),
- },
- .underlayer = &hash_rxq_init[HASH_RXQ_ETH],
- },
- [HASH_RXQ_TCPV6] = {
- .hash_fields = (IBV_RX_HASH_SRC_IPV6 |
- IBV_RX_HASH_DST_IPV6 |
- IBV_RX_HASH_SRC_PORT_TCP |
- IBV_RX_HASH_DST_PORT_TCP),
- .dpdk_rss_hf = ETH_RSS_NONFRAG_IPV6_TCP,
- .flow_priority = 0,
- .flow_spec.tcp_udp = {
- .type = IBV_FLOW_SPEC_TCP,
- .size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
- },
- .underlayer = &hash_rxq_init[HASH_RXQ_IPV6],
- },
- [HASH_RXQ_UDPV6] = {
- .hash_fields = (IBV_RX_HASH_SRC_IPV6 |
- IBV_RX_HASH_DST_IPV6 |
- IBV_RX_HASH_SRC_PORT_UDP |
- IBV_RX_HASH_DST_PORT_UDP),
- .dpdk_rss_hf = ETH_RSS_NONFRAG_IPV6_UDP,
- .flow_priority = 0,
- .flow_spec.tcp_udp = {
- .type = IBV_FLOW_SPEC_UDP,
- .size = sizeof(hash_rxq_init[0].flow_spec.tcp_udp),
- },
- .underlayer = &hash_rxq_init[HASH_RXQ_IPV6],
- },
- [HASH_RXQ_IPV6] = {
- .hash_fields = (IBV_RX_HASH_SRC_IPV6 |
- IBV_RX_HASH_DST_IPV6),
- .dpdk_rss_hf = (ETH_RSS_IPV6 |
- ETH_RSS_FRAG_IPV6),
- .flow_priority = 1,
- .flow_spec.ipv6 = {
- .type = IBV_FLOW_SPEC_IPV6,
- .size = sizeof(hash_rxq_init[0].flow_spec.ipv6),
- },
- .underlayer = &hash_rxq_init[HASH_RXQ_ETH],
- },
- [HASH_RXQ_ETH] = {
- .hash_fields = 0,
- .dpdk_rss_hf = 0,
- .flow_priority = 2,
- .flow_spec.eth = {
- .type = IBV_FLOW_SPEC_ETH,
- .size = sizeof(hash_rxq_init[0].flow_spec.eth),
- },
- .underlayer = NULL,
- },
-};
-/* Number of entries in hash_rxq_init[]. */
-const unsigned int hash_rxq_init_n = RTE_DIM(hash_rxq_init);
-
-/* Initialization data for hash RX queue indirection tables. */
-static const struct ind_table_init ind_table_init[] = {
- {
- .max_size = -1u, /* Superseded by HW limitations. */
- .hash_types =
- 1 << HASH_RXQ_TCPV4 |
- 1 << HASH_RXQ_UDPV4 |
- 1 << HASH_RXQ_IPV4 |
- 1 << HASH_RXQ_TCPV6 |
- 1 << HASH_RXQ_UDPV6 |
- 1 << HASH_RXQ_IPV6 |
- 0,
- .hash_types_n = 6,
- },
- {
- .max_size = 1,
- .hash_types = 1 << HASH_RXQ_ETH,
- .hash_types_n = 1,
- },
-};
-
-#define IND_TABLE_INIT_N RTE_DIM(ind_table_init)
/* Default RSS hash key also used for ConnectX-3. */
uint8_t rss_hash_default_key[] = {
};
/* Length of the default RSS hash key. */
-const size_t rss_hash_default_key_len = sizeof(rss_hash_default_key);
-
-/**
- * Populate flow steering rule for a given hash RX queue type using
- * information from hash_rxq_init[]. Nothing is written to flow_attr when
- * flow_attr_size is not large enough, but the required size is still returned.
- *
- * @param priv
- * Pointer to private structure.
- * @param[out] flow_attr
- * Pointer to flow attribute structure to fill. Note that the allocated
- * area must be larger and large enough to hold all flow specifications.
- * @param flow_attr_size
- * Entire size of flow_attr and trailing room for flow specifications.
- * @param type
- * Hash RX queue type to use for flow steering rule.
- *
- * @return
- * Total size of the flow attribute buffer. No errors are defined.
- */
-size_t
-priv_flow_attr(struct priv *priv, struct ibv_flow_attr *flow_attr,
- size_t flow_attr_size, enum hash_rxq_type type)
-{
- size_t offset = sizeof(*flow_attr);
- const struct hash_rxq_init *init = &hash_rxq_init[type];
-
- assert(priv != NULL);
- assert((size_t)type < RTE_DIM(hash_rxq_init));
- do {
- offset += init->flow_spec.hdr.size;
- init = init->underlayer;
- } while (init != NULL);
- if (offset > flow_attr_size)
- return offset;
- flow_attr_size = offset;
- init = &hash_rxq_init[type];
- *flow_attr = (struct ibv_flow_attr){
- .type = IBV_FLOW_ATTR_NORMAL,
- /* Priorities < 3 are reserved for flow director. */
- .priority = init->flow_priority + 3,
- .num_of_specs = 0,
- .port = priv->port,
- .flags = 0,
- };
- do {
- offset -= init->flow_spec.hdr.size;
- memcpy((void *)((uintptr_t)flow_attr + offset),
- &init->flow_spec,
- init->flow_spec.hdr.size);
- ++flow_attr->num_of_specs;
- init = init->underlayer;
- } while (init != NULL);
- return flow_attr_size;
-}
+static_assert(MLX5_RSS_HASH_KEY_LEN ==
+ (unsigned int)sizeof(rss_hash_default_key),
+ "wrong RSS default key size.");
/**
- * Convert hash type position in indirection table initializer to
- * hash RX queue type.
+ * Calculate the number of CQEs in CQ for the Rx queue.
*
- * @param table
- * Indirection table initializer.
- * @param pos
- * Hash type position.
+ * @param rxq_data
+ * Pointer to receive queue structure.
*
* @return
- * Hash RX queue type.
+ * Number of CQEs in CQ.
*/
-static enum hash_rxq_type
-hash_rxq_type_from_pos(const struct ind_table_init *table, unsigned int pos)
+unsigned int
+mlx5_rxq_cqe_num(struct mlx5_rxq_data *rxq_data)
{
- enum hash_rxq_type type = HASH_RXQ_TCPV4;
+ unsigned int cqe_n;
+ unsigned int wqe_n = 1 << rxq_data->elts_n;
- assert(pos < table->hash_types_n);
- do {
- if ((table->hash_types & (1 << type)) && (pos-- == 0))
- break;
- ++type;
- } while (1);
- return type;
+ if (mlx5_rxq_mprq_enabled(rxq_data))
+ cqe_n = wqe_n * (1 << rxq_data->strd_num_n) - 1;
+ else
+ cqe_n = wqe_n - 1;
+ return cqe_n;
}
/**
- * Filter out disabled hash RX queue types from ind_table_init[].
+ * Allocate RX queue elements for Multi-Packet RQ.
*
- * @param priv
- * Pointer to private structure.
- * @param[out] table
- * Output table.
+ * @param rxq_ctrl
+ * Pointer to RX queue structure.
*
* @return
- * Number of table entries.
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
*/
-static unsigned int
-priv_make_ind_table_init(struct priv *priv,
- struct ind_table_init (*table)[IND_TABLE_INIT_N])
+static int
+rxq_alloc_elts_mprq(struct mlx5_rxq_ctrl *rxq_ctrl)
{
- uint64_t rss_hf;
+ struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
+ unsigned int wqe_n = 1 << rxq->elts_n;
unsigned int i;
- unsigned int j;
- unsigned int table_n = 0;
- /* Mandatory to receive frames not handled by normal hash RX queues. */
- unsigned int hash_types_sup = 1 << HASH_RXQ_ETH;
-
- rss_hf = priv->rss_conf.rss_hf;
- /* Process other protocols only if more than one queue. */
- if (priv->rxqs_n > 1)
- for (i = 0; (i != hash_rxq_init_n); ++i)
- if (rss_hf & hash_rxq_init[i].dpdk_rss_hf)
- hash_types_sup |= (1 << i);
-
- /* Filter out entries whose protocols are not in the set. */
- for (i = 0, j = 0; (i != IND_TABLE_INIT_N); ++i) {
- unsigned int nb;
- unsigned int h;
-
- /* j is increased only if the table has valid protocols. */
- assert(j <= i);
- (*table)[j] = ind_table_init[i];
- (*table)[j].hash_types &= hash_types_sup;
- for (h = 0, nb = 0; (h != hash_rxq_init_n); ++h)
- if (((*table)[j].hash_types >> h) & 0x1)
- ++nb;
- (*table)[i].hash_types_n = nb;
- if (nb) {
- ++table_n;
- ++j;
- }
- }
- return table_n;
-}
+ int err;
-/**
- * Initialize hash RX queues and indirection table.
- *
- * @param priv
- * Pointer to private structure.
- *
- * @return
- * 0 on success, errno value on failure.
- */
-int
-priv_create_hash_rxqs(struct priv *priv)
-{
- struct ibv_wq *wqs[priv->reta_idx_n];
- struct ind_table_init ind_table_init[IND_TABLE_INIT_N];
- unsigned int ind_tables_n =
- priv_make_ind_table_init(priv, &ind_table_init);
- unsigned int hash_rxqs_n = 0;
- struct hash_rxq (*hash_rxqs)[] = NULL;
- struct ibv_rwq_ind_table *(*ind_tables)[] = NULL;
- unsigned int i;
- unsigned int j;
- unsigned int k;
- int err = 0;
-
- assert(priv->ind_tables == NULL);
- assert(priv->ind_tables_n == 0);
- assert(priv->hash_rxqs == NULL);
- assert(priv->hash_rxqs_n == 0);
- assert(priv->pd != NULL);
- assert(priv->ctx != NULL);
- if (priv->isolated)
- return 0;
- if (priv->rxqs_n == 0)
- return EINVAL;
- assert(priv->rxqs != NULL);
- if (ind_tables_n == 0) {
- ERROR("all hash RX queue types have been filtered out,"
- " indirection table cannot be created");
- return EINVAL;
- }
- if (priv->rxqs_n & (priv->rxqs_n - 1)) {
- INFO("%u RX queues are configured, consider rounding this"
- " number to the next power of two for better balancing",
- priv->rxqs_n);
- DEBUG("indirection table extended to assume %u WQs",
- priv->reta_idx_n);
- }
- for (i = 0; (i != priv->reta_idx_n); ++i) {
- struct mlx5_rxq_ctrl *rxq_ctrl;
-
- rxq_ctrl = container_of((*priv->rxqs)[(*priv->reta_idx)[i]],
- struct mlx5_rxq_ctrl, rxq);
- wqs[i] = rxq_ctrl->ibv->wq;
- }
- /* Get number of hash RX queues to configure. */
- for (i = 0, hash_rxqs_n = 0; (i != ind_tables_n); ++i)
- hash_rxqs_n += ind_table_init[i].hash_types_n;
- DEBUG("allocating %u hash RX queues for %u WQs, %u indirection tables",
- hash_rxqs_n, priv->rxqs_n, ind_tables_n);
- /* Create indirection tables. */
- ind_tables = rte_calloc(__func__, ind_tables_n,
- sizeof((*ind_tables)[0]), 0);
- if (ind_tables == NULL) {
- err = ENOMEM;
- ERROR("cannot allocate indirection tables container: %s",
- strerror(err));
- goto error;
- }
- for (i = 0; (i != ind_tables_n); ++i) {
- struct ibv_rwq_ind_table_init_attr ind_init_attr = {
- .log_ind_tbl_size = 0, /* Set below. */
- .ind_tbl = wqs,
- .comp_mask = 0,
- };
- unsigned int ind_tbl_size = ind_table_init[i].max_size;
- struct ibv_rwq_ind_table *ind_table;
-
- if (priv->reta_idx_n < ind_tbl_size)
- ind_tbl_size = priv->reta_idx_n;
- ind_init_attr.log_ind_tbl_size = log2above(ind_tbl_size);
- errno = 0;
- ind_table = ibv_create_rwq_ind_table(priv->ctx,
- &ind_init_attr);
- if (ind_table != NULL) {
- (*ind_tables)[i] = ind_table;
- continue;
- }
- /* Not clear whether errno is set. */
- err = (errno ? errno : EINVAL);
- ERROR("RX indirection table creation failed with error %d: %s",
- err, strerror(err));
- goto error;
- }
- /* Allocate array that holds hash RX queues and related data. */
- hash_rxqs = rte_calloc(__func__, hash_rxqs_n,
- sizeof((*hash_rxqs)[0]), 0);
- if (hash_rxqs == NULL) {
- err = ENOMEM;
- ERROR("cannot allocate hash RX queues container: %s",
- strerror(err));
- goto error;
- }
- for (i = 0, j = 0, k = 0;
- ((i != hash_rxqs_n) && (j != ind_tables_n));
- ++i) {
- struct hash_rxq *hash_rxq = &(*hash_rxqs)[i];
- enum hash_rxq_type type =
- hash_rxq_type_from_pos(&ind_table_init[j], k);
- struct rte_eth_rss_conf *priv_rss_conf = &priv->rss_conf;
- struct ibv_rx_hash_conf hash_conf = {
- .rx_hash_function = IBV_RX_HASH_FUNC_TOEPLITZ,
- .rx_hash_key_len = (priv_rss_conf ?
- priv_rss_conf->rss_key_len :
- rss_hash_default_key_len),
- .rx_hash_key = (priv_rss_conf ?
- priv_rss_conf->rss_key :
- rss_hash_default_key),
- .rx_hash_fields_mask = hash_rxq_init[type].hash_fields,
- };
- struct ibv_qp_init_attr_ex qp_init_attr = {
- .qp_type = IBV_QPT_RAW_PACKET,
- .comp_mask = (IBV_QP_INIT_ATTR_PD |
- IBV_QP_INIT_ATTR_IND_TABLE |
- IBV_QP_INIT_ATTR_RX_HASH),
- .rx_hash_conf = hash_conf,
- .rwq_ind_tbl = (*ind_tables)[j],
- .pd = priv->pd,
- };
-
- DEBUG("using indirection table %u for hash RX queue %u type %d",
- j, i, type);
- *hash_rxq = (struct hash_rxq){
- .priv = priv,
- .qp = ibv_create_qp_ex(priv->ctx, &qp_init_attr),
- .type = type,
- };
- if (hash_rxq->qp == NULL) {
- err = (errno ? errno : EINVAL);
- ERROR("Hash RX QP creation failure: %s",
- strerror(err));
+ /* Iterate on segments. */
+ for (i = 0; i <= wqe_n; ++i) {
+ struct mlx5_mprq_buf *buf;
+
+ if (rte_mempool_get(rxq->mprq_mp, (void **)&buf) < 0) {
+ DRV_LOG(ERR, "port %u empty mbuf pool", rxq->port_id);
+ rte_errno = ENOMEM;
goto error;
}
- if (++k < ind_table_init[j].hash_types_n)
- continue;
- /* Switch to the next indirection table and reset hash RX
- * queue type array index. */
- ++j;
- k = 0;
- }
- priv->ind_tables = ind_tables;
- priv->ind_tables_n = ind_tables_n;
- priv->hash_rxqs = hash_rxqs;
- priv->hash_rxqs_n = hash_rxqs_n;
- assert(err == 0);
+ if (i < wqe_n)
+ (*rxq->mprq_bufs)[i] = buf;
+ else
+ rxq->mprq_repl = buf;
+ }
+ DRV_LOG(DEBUG,
+ "port %u MPRQ queue %u allocated and configured %u segments",
+ rxq->port_id, rxq->idx, wqe_n);
return 0;
error:
- if (hash_rxqs != NULL) {
- for (i = 0; (i != hash_rxqs_n); ++i) {
- struct ibv_qp *qp = (*hash_rxqs)[i].qp;
-
- if (qp == NULL)
- continue;
- claim_zero(ibv_destroy_qp(qp));
- }
- rte_free(hash_rxqs);
- }
- if (ind_tables != NULL) {
- for (j = 0; (j != ind_tables_n); ++j) {
- struct ibv_rwq_ind_table *ind_table =
- (*ind_tables)[j];
-
- if (ind_table == NULL)
- continue;
- claim_zero(ibv_destroy_rwq_ind_table(ind_table));
- }
- rte_free(ind_tables);
- }
- return err;
-}
-
-/**
- * Clean up hash RX queues and indirection table.
- *
- * @param priv
- * Pointer to private structure.
- */
-void
-priv_destroy_hash_rxqs(struct priv *priv)
-{
- unsigned int i;
-
- DEBUG("destroying %u hash RX queues", priv->hash_rxqs_n);
- if (priv->hash_rxqs_n == 0) {
- assert(priv->hash_rxqs == NULL);
- assert(priv->ind_tables == NULL);
- return;
+ err = rte_errno; /* Save rte_errno before cleanup. */
+ wqe_n = i;
+ for (i = 0; (i != wqe_n); ++i) {
+ if ((*rxq->mprq_bufs)[i] != NULL)
+ rte_mempool_put(rxq->mprq_mp,
+ (*rxq->mprq_bufs)[i]);
+ (*rxq->mprq_bufs)[i] = NULL;
}
- for (i = 0; (i != priv->hash_rxqs_n); ++i) {
- struct hash_rxq *hash_rxq = &(*priv->hash_rxqs)[i];
- unsigned int j, k;
-
- assert(hash_rxq->priv == priv);
- assert(hash_rxq->qp != NULL);
- for (j = 0; (j != RTE_DIM(hash_rxq->mac_flow)); ++j)
- for (k = 0; (k != RTE_DIM(hash_rxq->mac_flow[j])); ++k)
- assert(hash_rxq->mac_flow[j][k] == NULL);
- claim_zero(ibv_destroy_qp(hash_rxq->qp));
- }
- priv->hash_rxqs_n = 0;
- rte_free(priv->hash_rxqs);
- priv->hash_rxqs = NULL;
- for (i = 0; (i != priv->ind_tables_n); ++i) {
- struct ibv_rwq_ind_table *ind_table =
- (*priv->ind_tables)[i];
-
- assert(ind_table != NULL);
- claim_zero(ibv_destroy_rwq_ind_table(ind_table));
- }
- priv->ind_tables_n = 0;
- rte_free(priv->ind_tables);
- priv->ind_tables = NULL;
+ DRV_LOG(DEBUG, "port %u MPRQ queue %u failed, freed everything",
+ rxq->port_id, rxq->idx);
+ rte_errno = err; /* Restore rte_errno. */
+ return -rte_errno;
}
/**
- * Allocate RX queue elements.
+ * Allocate RX queue elements for Single-Packet RQ.
*
* @param rxq_ctrl
* Pointer to RX queue structure.
* @return
* 0 on success, errno value on failure.
*/
-int
-rxq_alloc_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
+static int
+rxq_alloc_elts_sprq(struct mlx5_rxq_ctrl *rxq_ctrl)
{
const unsigned int sges_n = 1 << rxq_ctrl->rxq.sges_n;
- unsigned int elts_n = 1 << rxq_ctrl->rxq.elts_n;
+ unsigned int elts_n = mlx5_rxq_mprq_enabled(&rxq_ctrl->rxq) ?
+ (1 << rxq_ctrl->rxq.elts_n) * (1 << rxq_ctrl->rxq.strd_num_n) :
+ (1 << rxq_ctrl->rxq.elts_n);
unsigned int i;
- int ret = 0;
+ int err;
/* Iterate on segments. */
for (i = 0; (i != elts_n); ++i) {
+ struct mlx5_eth_rxseg *seg = &rxq_ctrl->rxq.rxseg[i % sges_n];
struct rte_mbuf *buf;
- buf = rte_pktmbuf_alloc(rxq_ctrl->rxq.mp);
+ buf = rte_pktmbuf_alloc(seg->mp);
if (buf == NULL) {
- ERROR("%p: empty mbuf pool", (void *)rxq_ctrl);
- ret = ENOMEM;
+ DRV_LOG(ERR, "port %u empty mbuf pool",
+ PORT_ID(rxq_ctrl->priv));
+ rte_errno = ENOMEM;
goto error;
}
/* Headroom is reserved by rte_pktmbuf_alloc(). */
- assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM);
+ MLX5_ASSERT(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM);
/* Buffer is supposed to be empty. */
- assert(rte_pktmbuf_data_len(buf) == 0);
- assert(rte_pktmbuf_pkt_len(buf) == 0);
- assert(!buf->next);
- /* Only the first segment keeps headroom. */
- if (i % sges_n)
- SET_DATA_OFF(buf, 0);
+ MLX5_ASSERT(rte_pktmbuf_data_len(buf) == 0);
+ MLX5_ASSERT(rte_pktmbuf_pkt_len(buf) == 0);
+ MLX5_ASSERT(!buf->next);
+ SET_DATA_OFF(buf, seg->offset);
PORT(buf) = rxq_ctrl->rxq.port_id;
- DATA_LEN(buf) = rte_pktmbuf_tailroom(buf);
- PKT_LEN(buf) = DATA_LEN(buf);
+ DATA_LEN(buf) = seg->length;
+ PKT_LEN(buf) = seg->length;
NB_SEGS(buf) = 1;
(*rxq_ctrl->rxq.elts)[i] = buf;
}
/* If Rx vector is activated. */
- if (rxq_check_vec_support(&rxq_ctrl->rxq) > 0) {
+ if (mlx5_rxq_check_vec_support(&rxq_ctrl->rxq) > 0) {
struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
struct rte_mbuf *mbuf_init = &rxq->fake_mbuf;
+ struct rte_pktmbuf_pool_private *priv =
+ (struct rte_pktmbuf_pool_private *)
+ rte_mempool_get_priv(rxq_ctrl->rxq.mp);
int j;
/* Initialize default rearm_data for vPMD. */
rte_mbuf_refcnt_set(mbuf_init, 1);
mbuf_init->nb_segs = 1;
mbuf_init->port = rxq->port_id;
+ if (priv->flags & RTE_PKTMBUF_POOL_F_PINNED_EXT_BUF)
+ mbuf_init->ol_flags = EXT_ATTACHED_MBUF;
/*
* prevent compiler reordering:
* rearm_data covers previous fields.
*/
rte_compiler_barrier();
rxq->mbuf_initializer =
- *(uint64_t *)&mbuf_init->rearm_data;
+ *(rte_xmm_t *)&mbuf_init->rearm_data;
/* Padding with a fake mbuf for vectorized Rx. */
for (j = 0; j < MLX5_VPMD_DESCS_PER_LOOP; ++j)
(*rxq->elts)[elts_n + j] = &rxq->fake_mbuf;
}
- DEBUG("%p: allocated and configured %u segments (max %u packets)",
- (void *)rxq_ctrl, elts_n, elts_n / (1 << rxq_ctrl->rxq.sges_n));
- assert(ret == 0);
+ DRV_LOG(DEBUG,
+ "port %u SPRQ queue %u allocated and configured %u segments"
+ " (max %u packets)",
+ PORT_ID(rxq_ctrl->priv), rxq_ctrl->rxq.idx, elts_n,
+ elts_n / (1 << rxq_ctrl->rxq.sges_n));
return 0;
error:
+ err = rte_errno; /* Save rte_errno before cleanup. */
elts_n = i;
for (i = 0; (i != elts_n); ++i) {
if ((*rxq_ctrl->rxq.elts)[i] != NULL)
rte_pktmbuf_free_seg((*rxq_ctrl->rxq.elts)[i]);
(*rxq_ctrl->rxq.elts)[i] = NULL;
}
- DEBUG("%p: failed, freed everything", (void *)rxq_ctrl);
- assert(ret > 0);
- return ret;
+ DRV_LOG(DEBUG, "port %u SPRQ queue %u failed, freed everything",
+ PORT_ID(rxq_ctrl->priv), rxq_ctrl->rxq.idx);
+ rte_errno = err; /* Restore rte_errno. */
+ return -rte_errno;
}
/**
- * Free RX queue elements.
+ * Allocate RX queue elements.
+ *
+ * @param rxq_ctrl
+ * Pointer to RX queue structure.
+ *
+ * @return
+ * 0 on success, errno value on failure.
+ */
+int
+rxq_alloc_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
+{
+ int ret = 0;
+
+ /**
+ * For MPRQ we need to allocate both MPRQ buffers
+ * for WQEs and simple mbufs for vector processing.
+ */
+ if (mlx5_rxq_mprq_enabled(&rxq_ctrl->rxq))
+ ret = rxq_alloc_elts_mprq(rxq_ctrl);
+ return (ret || rxq_alloc_elts_sprq(rxq_ctrl));
+}
+
+/**
+ * Free RX queue elements for Multi-Packet RQ.
*
* @param rxq_ctrl
* Pointer to RX queue structure.
*/
static void
-rxq_free_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
+rxq_free_elts_mprq(struct mlx5_rxq_ctrl *rxq_ctrl)
{
struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
- const uint16_t q_n = (1 << rxq->elts_n);
+ uint16_t i;
+
+ DRV_LOG(DEBUG, "port %u Multi-Packet Rx queue %u freeing %d WRs",
+ rxq->port_id, rxq->idx, (1u << rxq->elts_n));
+ if (rxq->mprq_bufs == NULL)
+ return;
+ for (i = 0; (i != (1u << rxq->elts_n)); ++i) {
+ if ((*rxq->mprq_bufs)[i] != NULL)
+ mlx5_mprq_buf_free((*rxq->mprq_bufs)[i]);
+ (*rxq->mprq_bufs)[i] = NULL;
+ }
+ if (rxq->mprq_repl != NULL) {
+ mlx5_mprq_buf_free(rxq->mprq_repl);
+ rxq->mprq_repl = NULL;
+ }
+}
+
+/**
+ * Free RX queue elements for Single-Packet RQ.
+ *
+ * @param rxq_ctrl
+ * Pointer to RX queue structure.
+ */
+static void
+rxq_free_elts_sprq(struct mlx5_rxq_ctrl *rxq_ctrl)
+{
+ struct mlx5_rxq_data *rxq = &rxq_ctrl->rxq;
+ const uint16_t q_n = mlx5_rxq_mprq_enabled(&rxq_ctrl->rxq) ?
+ (1 << rxq->elts_n) * (1 << rxq->strd_num_n) :
+ (1 << rxq->elts_n);
const uint16_t q_mask = q_n - 1;
- uint16_t used = q_n - (rxq->rq_ci - rxq->rq_pi);
+ uint16_t elts_ci = mlx5_rxq_mprq_enabled(&rxq_ctrl->rxq) ?
+ rxq->elts_ci : rxq->rq_ci;
+ uint16_t used = q_n - (elts_ci - rxq->rq_pi);
uint16_t i;
- DEBUG("%p: freeing WRs", (void *)rxq_ctrl);
+ DRV_LOG(DEBUG, "port %u Rx queue %u freeing %d WRs",
+ PORT_ID(rxq_ctrl->priv), rxq->idx, q_n);
if (rxq->elts == NULL)
return;
/**
- * Some mbuf in the Ring belongs to the application. They cannot be
- * freed.
+ * Some mbuf in the Ring belongs to the application.
+ * They cannot be freed.
*/
- if (rxq_check_vec_support(rxq) > 0) {
+ if (mlx5_rxq_check_vec_support(rxq) > 0) {
for (i = 0; i < used; ++i)
- (*rxq->elts)[(rxq->rq_ci + i) & q_mask] = NULL;
- rxq->rq_pi = rxq->rq_ci;
+ (*rxq->elts)[(elts_ci + i) & q_mask] = NULL;
+ rxq->rq_pi = elts_ci;
}
- for (i = 0; (i != (1u << rxq->elts_n)); ++i) {
+ for (i = 0; i != q_n; ++i) {
if ((*rxq->elts)[i] != NULL)
rte_pktmbuf_free_seg((*rxq->elts)[i]);
(*rxq->elts)[i] = NULL;
}
/**
- * Clean up a RX queue.
- *
- * Destroy objects, free allocated memory and reset the structure for reuse.
+ * Free RX queue elements.
*
* @param rxq_ctrl
* Pointer to RX queue structure.
*/
-void
-mlx5_rxq_cleanup(struct mlx5_rxq_ctrl *rxq_ctrl)
+static void
+rxq_free_elts(struct mlx5_rxq_ctrl *rxq_ctrl)
{
- DEBUG("cleaning up %p", (void *)rxq_ctrl);
- if (rxq_ctrl->ibv)
- mlx5_priv_rxq_ibv_release(rxq_ctrl->priv, rxq_ctrl->ibv);
- memset(rxq_ctrl, 0, sizeof(*rxq_ctrl));
+ /*
+ * For MPRQ we need to allocate both MPRQ buffers
+ * for WQEs and simple mbufs for vector processing.
+ */
+ if (mlx5_rxq_mprq_enabled(&rxq_ctrl->rxq))
+ rxq_free_elts_mprq(rxq_ctrl);
+ rxq_free_elts_sprq(rxq_ctrl);
}
/**
+ * Returns the per-queue supported offloads.
*
* @param dev
- * Pointer to Ethernet device structure.
- * @param idx
- * RX queue index.
- * @param desc
- * Number of descriptors to configure in queue.
- * @param socket
- * NUMA socket on which memory must be allocated.
- * @param[in] conf
- * Thresholds parameters.
- * @param mp
- * Memory pool for buffer allocations.
+ * Pointer to Ethernet device.
*
* @return
- * 0 on success, negative errno value on failure.
+ * Supported Rx offloads.
*/
-int
-mlx5_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
- unsigned int socket, const struct rte_eth_rxconf *conf,
- struct rte_mempool *mp)
+uint64_t
+mlx5_get_rx_queue_offloads(struct rte_eth_dev *dev)
{
- struct priv *priv = dev->data->dev_private;
- struct mlx5_rxq_data *rxq = (*priv->rxqs)[idx];
- struct mlx5_rxq_ctrl *rxq_ctrl =
- container_of(rxq, struct mlx5_rxq_ctrl, rxq);
- int ret = 0;
-
- (void)conf;
- if (mlx5_is_secondary())
- return -E_RTE_SECONDARY;
- priv_lock(priv);
- if (!rte_is_power_of_2(desc)) {
- desc = 1 << log2above(desc);
- WARN("%p: increased number of descriptors in RX queue %u"
- " to the next power of two (%d)",
- (void *)dev, idx, desc);
- }
- DEBUG("%p: configuring queue %u for %u descriptors",
- (void *)dev, idx, desc);
- if (idx >= priv->rxqs_n) {
- ERROR("%p: queue index out of range (%u >= %u)",
- (void *)dev, idx, priv->rxqs_n);
- priv_unlock(priv);
- return -EOVERFLOW;
- }
- if (!mlx5_priv_rxq_releasable(priv, idx)) {
- ret = EBUSY;
- ERROR("%p: unable to release queue index %u",
- (void *)dev, idx);
- goto out;
- }
- mlx5_priv_rxq_release(priv, idx);
- rxq_ctrl = mlx5_priv_rxq_new(priv, idx, desc, socket, mp);
- if (!rxq_ctrl) {
- ERROR("%p: unable to allocate queue index %u",
- (void *)dev, idx);
- ret = ENOMEM;
- goto out;
- }
- DEBUG("%p: adding RX queue %p to list",
- (void *)dev, (void *)rxq_ctrl);
- (*priv->rxqs)[idx] = &rxq_ctrl->rxq;
-out:
- priv_unlock(priv);
- return -ret;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_dev_config *config = &priv->config;
+ uint64_t offloads = (DEV_RX_OFFLOAD_SCATTER |
+ DEV_RX_OFFLOAD_TIMESTAMP |
+ DEV_RX_OFFLOAD_JUMBO_FRAME |
+ DEV_RX_OFFLOAD_RSS_HASH);
+
+ if (!config->mprq.enabled)
+ offloads |= RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT;
+ if (config->hw_fcs_strip)
+ offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
+ if (config->hw_csum)
+ offloads |= (DEV_RX_OFFLOAD_IPV4_CKSUM |
+ DEV_RX_OFFLOAD_UDP_CKSUM |
+ DEV_RX_OFFLOAD_TCP_CKSUM);
+ if (config->hw_vlan_strip)
+ offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
+ if (MLX5_LRO_SUPPORTED(dev))
+ offloads |= DEV_RX_OFFLOAD_TCP_LRO;
+ return offloads;
}
+
/**
- * DPDK callback to release a RX queue.
+ * Returns the per-port supported offloads.
*
- * @param dpdk_rxq
- * Generic RX queue pointer.
+ * @return
+ * Supported Rx offloads.
*/
-void
-mlx5_rx_queue_release(void *dpdk_rxq)
+uint64_t
+mlx5_get_rx_port_offloads(void)
{
- struct mlx5_rxq_data *rxq = (struct mlx5_rxq_data *)dpdk_rxq;
- struct mlx5_rxq_ctrl *rxq_ctrl;
- struct priv *priv;
-
- if (mlx5_is_secondary())
- return;
+ uint64_t offloads = DEV_RX_OFFLOAD_VLAN_FILTER;
- if (rxq == NULL)
- return;
- rxq_ctrl = container_of(rxq, struct mlx5_rxq_ctrl, rxq);
- priv = rxq_ctrl->priv;
- priv_lock(priv);
- if (!mlx5_priv_rxq_releasable(priv, rxq_ctrl->rxq.stats.idx))
- rte_panic("Rx queue %p is still used by a flow and cannot be"
- " removed\n", (void *)rxq_ctrl);
- mlx5_priv_rxq_release(priv, rxq_ctrl->rxq.stats.idx);
- priv_unlock(priv);
+ return offloads;
}
/**
- * Allocate queue vector and fill epoll fd list for Rx interrupts.
+ * Verify if the queue can be released.
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * RX queue index.
*
* @return
- * 0 on success, negative on failure.
+ * 1 if the queue can be released
+ * 0 if the queue can not be released, there are references to it.
+ * Negative errno and rte_errno is set if queue doesn't exist.
*/
-int
-priv_rx_intr_vec_enable(struct priv *priv)
+static int
+mlx5_rxq_releasable(struct rte_eth_dev *dev, uint16_t idx)
{
- unsigned int i;
- unsigned int rxqs_n = priv->rxqs_n;
- unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
- unsigned int count = 0;
- struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
-
- assert(!mlx5_is_secondary());
- if (!priv->dev->data->dev_conf.intr_conf.rxq)
- return 0;
- priv_rx_intr_vec_disable(priv);
- intr_handle->intr_vec = malloc(sizeof(intr_handle->intr_vec[rxqs_n]));
- if (intr_handle->intr_vec == NULL) {
- ERROR("failed to allocate memory for interrupt vector,"
- " Rx interrupts will not be supported");
- return -ENOMEM;
- }
- intr_handle->type = RTE_INTR_HANDLE_EXT;
- for (i = 0; i != n; ++i) {
- /* This rxq ibv must not be released in this function. */
- struct mlx5_rxq_ibv *rxq_ibv = mlx5_priv_rxq_ibv_get(priv, i);
- int fd;
- int flags;
- int rc;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *rxq_ctrl;
- /* Skip queues that cannot request interrupts. */
- if (!rxq_ibv || !rxq_ibv->channel) {
- /* Use invalid intr_vec[] index to disable entry. */
- intr_handle->intr_vec[i] =
- RTE_INTR_VEC_RXTX_OFFSET +
- RTE_MAX_RXTX_INTR_VEC_ID;
- continue;
- }
- if (count >= RTE_MAX_RXTX_INTR_VEC_ID) {
- ERROR("too many Rx queues for interrupt vector size"
- " (%d), Rx interrupts cannot be enabled",
- RTE_MAX_RXTX_INTR_VEC_ID);
- priv_rx_intr_vec_disable(priv);
- return -1;
- }
- fd = rxq_ibv->channel->fd;
- flags = fcntl(fd, F_GETFL);
- rc = fcntl(fd, F_SETFL, flags | O_NONBLOCK);
- if (rc < 0) {
- ERROR("failed to make Rx interrupt file descriptor"
- " %d non-blocking for queue index %d", fd, i);
- priv_rx_intr_vec_disable(priv);
- return -1;
- }
- intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + count;
- intr_handle->efds[count] = fd;
- count++;
+ if (!(*priv->rxqs)[idx]) {
+ rte_errno = EINVAL;
+ return -rte_errno;
}
- if (!count)
- priv_rx_intr_vec_disable(priv);
- else
- intr_handle->nb_efd = count;
- return 0;
+ rxq_ctrl = container_of((*priv->rxqs)[idx], struct mlx5_rxq_ctrl, rxq);
+ return (__atomic_load_n(&rxq_ctrl->refcnt, __ATOMIC_RELAXED) == 1);
}
-/**
- * Clean up Rx interrupts handler.
- *
- * @param priv
- * Pointer to private structure.
- */
-void
-priv_rx_intr_vec_disable(struct priv *priv)
+/* Fetches and drops all SW-owned and error CQEs to synchronize CQ. */
+static void
+rxq_sync_cq(struct mlx5_rxq_data *rxq)
{
- struct rte_intr_handle *intr_handle = priv->dev->intr_handle;
- unsigned int i;
- unsigned int rxqs_n = priv->rxqs_n;
- unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
-
- if (!priv->dev->data->dev_conf.intr_conf.rxq)
- return;
- for (i = 0; i != n; ++i) {
- struct mlx5_rxq_ctrl *rxq_ctrl;
- struct mlx5_rxq_data *rxq_data;
+ const uint16_t cqe_n = 1 << rxq->cqe_n;
+ const uint16_t cqe_mask = cqe_n - 1;
+ volatile struct mlx5_cqe *cqe;
+ int ret, i;
- if (intr_handle->intr_vec[i] == RTE_INTR_VEC_RXTX_OFFSET +
- RTE_MAX_RXTX_INTR_VEC_ID)
+ i = cqe_n;
+ do {
+ cqe = &(*rxq->cqes)[rxq->cq_ci & cqe_mask];
+ ret = check_cqe(cqe, cqe_n, rxq->cq_ci);
+ if (ret == MLX5_CQE_STATUS_HW_OWN)
+ break;
+ if (ret == MLX5_CQE_STATUS_ERR) {
+ rxq->cq_ci++;
continue;
- /**
- * Need to access directly the queue to release the reference
- * kept in priv_rx_intr_vec_enable().
- */
- rxq_data = (*priv->rxqs)[i];
- rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
- mlx5_priv_rxq_ibv_release(priv, rxq_ctrl->ibv);
+ }
+ MLX5_ASSERT(ret == MLX5_CQE_STATUS_SW_OWN);
+ if (MLX5_CQE_FORMAT(cqe->op_own) != MLX5_COMPRESSED) {
+ rxq->cq_ci++;
+ continue;
+ }
+ /* Compute the next non compressed CQE. */
+ rxq->cq_ci += rte_be_to_cpu_32(cqe->byte_cnt);
+
+ } while (--i);
+ /* Move all CQEs to HW ownership, including possible MiniCQEs. */
+ for (i = 0; i < cqe_n; i++) {
+ cqe = &(*rxq->cqes)[i];
+ cqe->op_own = MLX5_CQE_INVALIDATE;
}
- rte_intr_free_epoll_fd(intr_handle);
- free(intr_handle->intr_vec);
- intr_handle->nb_efd = 0;
- intr_handle->intr_vec = NULL;
+ /* Resync CQE and WQE (WQ in RESET state). */
+ rte_io_wmb();
+ *rxq->cq_db = rte_cpu_to_be_32(rxq->cq_ci);
+ rte_io_wmb();
+ *rxq->rq_db = rte_cpu_to_be_32(0);
+ rte_io_wmb();
}
/**
- * MLX5 CQ notification .
+ * Rx queue stop. Device queue goes to the RESET state,
+ * all involved mbufs are freed from WQ.
*
- * @param rxq
- * Pointer to receive queue structure.
- * @param sq_n_rxq
- * Sequence number per receive queue .
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param idx
+ * RX queue index.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
*/
-static inline void
-mlx5_arm_cq(struct mlx5_rxq_data *rxq, int sq_n_rxq)
+int
+mlx5_rx_queue_stop_primary(struct rte_eth_dev *dev, uint16_t idx)
{
- int sq_n = 0;
- uint32_t doorbell_hi;
- uint64_t doorbell;
- void *cq_db_reg = (char *)rxq->cq_uar + MLX5_CQ_DOORBELL;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[idx];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ int ret;
- sq_n = sq_n_rxq & MLX5_CQ_SQN_MASK;
- doorbell_hi = sq_n << MLX5_CQ_SQN_OFFSET | (rxq->cq_ci & MLX5_CI_MASK);
- doorbell = (uint64_t)doorbell_hi << 32;
- doorbell |= rxq->cqn;
- rxq->cq_db[MLX5_CQ_ARM_DB] = rte_cpu_to_be_32(doorbell_hi);
- rte_wmb();
- rte_write64(rte_cpu_to_be_64(doorbell), cq_db_reg);
+ MLX5_ASSERT(rte_eal_process_type() == RTE_PROC_PRIMARY);
+ ret = priv->obj_ops.rxq_obj_modify(rxq_ctrl->obj, MLX5_RXQ_MOD_RDY2RST);
+ if (ret) {
+ DRV_LOG(ERR, "Cannot change Rx WQ state to RESET: %s",
+ strerror(errno));
+ rte_errno = errno;
+ return ret;
+ }
+ /* Remove all processes CQEs. */
+ rxq_sync_cq(rxq);
+ /* Free all involved mbufs. */
+ rxq_free_elts(rxq_ctrl);
+ /* Set the actual queue state. */
+ dev->data->rx_queue_state[idx] = RTE_ETH_QUEUE_STATE_STOPPED;
+ return 0;
}
/**
- * DPDK callback for Rx queue interrupt enable.
+ * Rx queue stop. Device queue goes to the RESET state,
+ * all involved mbufs are freed from WQ.
*
* @param dev
* Pointer to Ethernet device structure.
- * @param rx_queue_id
- * Rx queue number.
+ * @param idx
+ * RX queue index.
*
* @return
- * 0 on success, negative on failure.
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
*/
int
-mlx5_rx_intr_enable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
+mlx5_rx_queue_stop(struct rte_eth_dev *dev, uint16_t idx)
{
- struct priv *priv = mlx5_get_priv(dev);
- struct mlx5_rxq_data *rxq_data;
+ eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
+ int ret;
+
+ if (rte_eth_dev_is_rx_hairpin_queue(dev, idx)) {
+ DRV_LOG(ERR, "Hairpin queue can't be stopped");
+ rte_errno = EINVAL;
+ return -EINVAL;
+ }
+ if (dev->data->rx_queue_state[idx] == RTE_ETH_QUEUE_STATE_STOPPED)
+ return 0;
+ /*
+ * Vectorized Rx burst requires the CQ and RQ indices
+ * synchronized, that might be broken on RQ restart
+ * and cause Rx malfunction, so queue stopping is
+ * not supported if vectorized Rx burst is engaged.
+ * The routine pointer depends on the process
+ * type, should perform check there.
+ */
+ if (pkt_burst == mlx5_rx_burst_vec) {
+ DRV_LOG(ERR, "Rx queue stop is not supported "
+ "for vectorized Rx");
+ rte_errno = EINVAL;
+ return -EINVAL;
+ }
+ if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
+ ret = mlx5_mp_os_req_queue_control(dev, idx,
+ MLX5_MP_REQ_QUEUE_RX_STOP);
+ } else {
+ ret = mlx5_rx_queue_stop_primary(dev, idx);
+ }
+ return ret;
+}
+
+/**
+ * Rx queue start. Device queue goes to the ready state,
+ * all required mbufs are allocated and WQ is replenished.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param idx
+ * RX queue index.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_rx_queue_start_primary(struct rte_eth_dev *dev, uint16_t idx)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[idx];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ int ret;
+
+ MLX5_ASSERT(rte_eal_process_type() == RTE_PROC_PRIMARY);
+ /* Allocate needed buffers. */
+ ret = rxq_alloc_elts(rxq_ctrl);
+ if (ret) {
+ DRV_LOG(ERR, "Cannot reallocate buffers for Rx WQ");
+ rte_errno = errno;
+ return ret;
+ }
+ rte_io_wmb();
+ *rxq->cq_db = rte_cpu_to_be_32(rxq->cq_ci);
+ rte_io_wmb();
+ /* Reset RQ consumer before moving queue to READY state. */
+ *rxq->rq_db = rte_cpu_to_be_32(0);
+ rte_io_wmb();
+ ret = priv->obj_ops.rxq_obj_modify(rxq_ctrl->obj, MLX5_RXQ_MOD_RST2RDY);
+ if (ret) {
+ DRV_LOG(ERR, "Cannot change Rx WQ state to READY: %s",
+ strerror(errno));
+ rte_errno = errno;
+ return ret;
+ }
+ /* Reinitialize RQ - set WQEs. */
+ mlx5_rxq_initialize(rxq);
+ rxq->err_state = MLX5_RXQ_ERR_STATE_NO_ERROR;
+ /* Set actual queue state. */
+ dev->data->rx_queue_state[idx] = RTE_ETH_QUEUE_STATE_STARTED;
+ return 0;
+}
+
+/**
+ * Rx queue start. Device queue goes to the ready state,
+ * all required mbufs are allocated and WQ is replenished.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param idx
+ * RX queue index.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_rx_queue_start(struct rte_eth_dev *dev, uint16_t idx)
+{
+ int ret;
+
+ if (rte_eth_dev_is_rx_hairpin_queue(dev, idx)) {
+ DRV_LOG(ERR, "Hairpin queue can't be started");
+ rte_errno = EINVAL;
+ return -EINVAL;
+ }
+ if (dev->data->rx_queue_state[idx] == RTE_ETH_QUEUE_STATE_STARTED)
+ return 0;
+ if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
+ ret = mlx5_mp_os_req_queue_control(dev, idx,
+ MLX5_MP_REQ_QUEUE_RX_START);
+ } else {
+ ret = mlx5_rx_queue_start_primary(dev, idx);
+ }
+ return ret;
+}
+
+/**
+ * Rx queue presetup checks.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param idx
+ * RX queue index.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_rx_queue_pre_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t *desc)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+
+ if (!rte_is_power_of_2(*desc)) {
+ *desc = 1 << log2above(*desc);
+ DRV_LOG(WARNING,
+ "port %u increased number of descriptors in Rx queue %u"
+ " to the next power of two (%d)",
+ dev->data->port_id, idx, *desc);
+ }
+ DRV_LOG(DEBUG, "port %u configuring Rx queue %u for %u descriptors",
+ dev->data->port_id, idx, *desc);
+ if (idx >= priv->rxqs_n) {
+ DRV_LOG(ERR, "port %u Rx queue index out of range (%u >= %u)",
+ dev->data->port_id, idx, priv->rxqs_n);
+ rte_errno = EOVERFLOW;
+ return -rte_errno;
+ }
+ if (!mlx5_rxq_releasable(dev, idx)) {
+ DRV_LOG(ERR, "port %u unable to release queue index %u",
+ dev->data->port_id, idx);
+ rte_errno = EBUSY;
+ return -rte_errno;
+ }
+ mlx5_rxq_release(dev, idx);
+ return 0;
+}
+
+/**
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param idx
+ * RX queue index.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param socket
+ * NUMA socket on which memory must be allocated.
+ * @param[in] conf
+ * Thresholds parameters.
+ * @param mp
+ * Memory pool for buffer allocations.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
+ unsigned int socket, const struct rte_eth_rxconf *conf,
+ struct rte_mempool *mp)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[idx];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ struct rte_eth_rxseg_split *rx_seg =
+ (struct rte_eth_rxseg_split *)conf->rx_seg;
+ struct rte_eth_rxseg_split rx_single = {.mp = mp};
+ uint16_t n_seg = conf->rx_nseg;
+ int res;
+
+ if (mp) {
+ /*
+ * The parameters should be checked on rte_eth_dev layer.
+ * If mp is specified it means the compatible configuration
+ * without buffer split feature tuning.
+ */
+ rx_seg = &rx_single;
+ n_seg = 1;
+ }
+ if (n_seg > 1) {
+ uint64_t offloads = conf->offloads |
+ dev->data->dev_conf.rxmode.offloads;
+
+ /* The offloads should be checked on rte_eth_dev layer. */
+ MLX5_ASSERT(offloads & DEV_RX_OFFLOAD_SCATTER);
+ if (!(offloads & RTE_ETH_RX_OFFLOAD_BUFFER_SPLIT)) {
+ DRV_LOG(ERR, "port %u queue index %u split "
+ "offload not configured",
+ dev->data->port_id, idx);
+ rte_errno = ENOSPC;
+ return -rte_errno;
+ }
+ MLX5_ASSERT(n_seg < MLX5_MAX_RXQ_NSEG);
+ }
+ res = mlx5_rx_queue_pre_setup(dev, idx, &desc);
+ if (res)
+ return res;
+ rxq_ctrl = mlx5_rxq_new(dev, idx, desc, socket, conf, rx_seg, n_seg);
+ if (!rxq_ctrl) {
+ DRV_LOG(ERR, "port %u unable to allocate queue index %u",
+ dev->data->port_id, idx);
+ rte_errno = ENOMEM;
+ return -rte_errno;
+ }
+ DRV_LOG(DEBUG, "port %u adding Rx queue %u to list",
+ dev->data->port_id, idx);
+ (*priv->rxqs)[idx] = &rxq_ctrl->rxq;
+ return 0;
+}
+
+/**
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param idx
+ * RX queue index.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param hairpin_conf
+ * Hairpin configuration parameters.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_rx_hairpin_queue_setup(struct rte_eth_dev *dev, uint16_t idx,
+ uint16_t desc,
+ const struct rte_eth_hairpin_conf *hairpin_conf)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[idx];
+ struct mlx5_rxq_ctrl *rxq_ctrl =
+ container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ int res;
+
+ res = mlx5_rx_queue_pre_setup(dev, idx, &desc);
+ if (res)
+ return res;
+ if (hairpin_conf->peer_count != 1) {
+ rte_errno = EINVAL;
+ DRV_LOG(ERR, "port %u unable to setup Rx hairpin queue index %u"
+ " peer count is %u", dev->data->port_id,
+ idx, hairpin_conf->peer_count);
+ return -rte_errno;
+ }
+ if (hairpin_conf->peers[0].port == dev->data->port_id) {
+ if (hairpin_conf->peers[0].queue >= priv->txqs_n) {
+ rte_errno = EINVAL;
+ DRV_LOG(ERR, "port %u unable to setup Rx hairpin queue"
+ " index %u, Tx %u is larger than %u",
+ dev->data->port_id, idx,
+ hairpin_conf->peers[0].queue, priv->txqs_n);
+ return -rte_errno;
+ }
+ } else {
+ if (hairpin_conf->manual_bind == 0 ||
+ hairpin_conf->tx_explicit == 0) {
+ rte_errno = EINVAL;
+ DRV_LOG(ERR, "port %u unable to setup Rx hairpin queue"
+ " index %u peer port %u with attributes %u %u",
+ dev->data->port_id, idx,
+ hairpin_conf->peers[0].port,
+ hairpin_conf->manual_bind,
+ hairpin_conf->tx_explicit);
+ return -rte_errno;
+ }
+ }
+ rxq_ctrl = mlx5_rxq_hairpin_new(dev, idx, desc, hairpin_conf);
+ if (!rxq_ctrl) {
+ DRV_LOG(ERR, "port %u unable to allocate queue index %u",
+ dev->data->port_id, idx);
+ rte_errno = ENOMEM;
+ return -rte_errno;
+ }
+ DRV_LOG(DEBUG, "port %u adding Rx queue %u to list",
+ dev->data->port_id, idx);
+ (*priv->rxqs)[idx] = &rxq_ctrl->rxq;
+ return 0;
+}
+
+/**
+ * DPDK callback to release a RX queue.
+ *
+ * @param dpdk_rxq
+ * Generic RX queue pointer.
+ */
+void
+mlx5_rx_queue_release(void *dpdk_rxq)
+{
+ struct mlx5_rxq_data *rxq = (struct mlx5_rxq_data *)dpdk_rxq;
+ struct mlx5_rxq_ctrl *rxq_ctrl;
+ struct mlx5_priv *priv;
+
+ if (rxq == NULL)
+ return;
+ rxq_ctrl = container_of(rxq, struct mlx5_rxq_ctrl, rxq);
+ priv = rxq_ctrl->priv;
+ if (!mlx5_rxq_releasable(ETH_DEV(priv), rxq_ctrl->rxq.idx))
+ rte_panic("port %u Rx queue %u is still used by a flow and"
+ " cannot be removed\n",
+ PORT_ID(priv), rxq->idx);
+ mlx5_rxq_release(ETH_DEV(priv), rxq_ctrl->rxq.idx);
+}
+
+/**
+ * Allocate queue vector and fill epoll fd list for Rx interrupts.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_rx_intr_vec_enable(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ unsigned int i;
+ unsigned int rxqs_n = priv->rxqs_n;
+ unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
+ unsigned int count = 0;
+ struct rte_intr_handle *intr_handle = dev->intr_handle;
+
+ /* Representor shares dev->intr_handle with PF. */
+ if (priv->representor)
+ return 0;
+ if (!dev->data->dev_conf.intr_conf.rxq)
+ return 0;
+ mlx5_rx_intr_vec_disable(dev);
+ intr_handle->intr_vec = mlx5_malloc(0,
+ n * sizeof(intr_handle->intr_vec[0]),
+ 0, SOCKET_ID_ANY);
+ if (intr_handle->intr_vec == NULL) {
+ DRV_LOG(ERR,
+ "port %u failed to allocate memory for interrupt"
+ " vector, Rx interrupts will not be supported",
+ dev->data->port_id);
+ rte_errno = ENOMEM;
+ return -rte_errno;
+ }
+ intr_handle->type = RTE_INTR_HANDLE_EXT;
+ for (i = 0; i != n; ++i) {
+ /* This rxq obj must not be released in this function. */
+ struct mlx5_rxq_ctrl *rxq_ctrl = mlx5_rxq_get(dev, i);
+ struct mlx5_rxq_obj *rxq_obj = rxq_ctrl ? rxq_ctrl->obj : NULL;
+ int rc;
+
+ /* Skip queues that cannot request interrupts. */
+ if (!rxq_obj || (!rxq_obj->ibv_channel &&
+ !rxq_obj->devx_channel)) {
+ /* Use invalid intr_vec[] index to disable entry. */
+ intr_handle->intr_vec[i] =
+ RTE_INTR_VEC_RXTX_OFFSET +
+ RTE_MAX_RXTX_INTR_VEC_ID;
+ /* Decrease the rxq_ctrl's refcnt */
+ if (rxq_ctrl)
+ mlx5_rxq_release(dev, i);
+ continue;
+ }
+ if (count >= RTE_MAX_RXTX_INTR_VEC_ID) {
+ DRV_LOG(ERR,
+ "port %u too many Rx queues for interrupt"
+ " vector size (%d), Rx interrupts cannot be"
+ " enabled",
+ dev->data->port_id, RTE_MAX_RXTX_INTR_VEC_ID);
+ mlx5_rx_intr_vec_disable(dev);
+ rte_errno = ENOMEM;
+ return -rte_errno;
+ }
+ rc = mlx5_os_set_nonblock_channel_fd(rxq_obj->fd);
+ if (rc < 0) {
+ rte_errno = errno;
+ DRV_LOG(ERR,
+ "port %u failed to make Rx interrupt file"
+ " descriptor %d non-blocking for queue index"
+ " %d",
+ dev->data->port_id, rxq_obj->fd, i);
+ mlx5_rx_intr_vec_disable(dev);
+ return -rte_errno;
+ }
+ intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + count;
+ intr_handle->efds[count] = rxq_obj->fd;
+ count++;
+ }
+ if (!count)
+ mlx5_rx_intr_vec_disable(dev);
+ else
+ intr_handle->nb_efd = count;
+ return 0;
+}
+
+/**
+ * Clean up Rx interrupts handler.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ */
+void
+mlx5_rx_intr_vec_disable(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct rte_intr_handle *intr_handle = dev->intr_handle;
+ unsigned int i;
+ unsigned int rxqs_n = priv->rxqs_n;
+ unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID);
+
+ /* Representor shares dev->intr_handle with PF. */
+ if (priv->representor)
+ return;
+ if (!dev->data->dev_conf.intr_conf.rxq)
+ return;
+ if (!intr_handle->intr_vec)
+ goto free;
+ for (i = 0; i != n; ++i) {
+ if (intr_handle->intr_vec[i] == RTE_INTR_VEC_RXTX_OFFSET +
+ RTE_MAX_RXTX_INTR_VEC_ID)
+ continue;
+ /**
+ * Need to access directly the queue to release the reference
+ * kept in mlx5_rx_intr_vec_enable().
+ */
+ mlx5_rxq_release(dev, i);
+ }
+free:
+ rte_intr_free_epoll_fd(intr_handle);
+ if (intr_handle->intr_vec)
+ mlx5_free(intr_handle->intr_vec);
+ intr_handle->nb_efd = 0;
+ intr_handle->intr_vec = NULL;
+}
+
+/**
+ * MLX5 CQ notification .
+ *
+ * @param rxq
+ * Pointer to receive queue structure.
+ * @param sq_n_rxq
+ * Sequence number per receive queue .
+ */
+static inline void
+mlx5_arm_cq(struct mlx5_rxq_data *rxq, int sq_n_rxq)
+{
+ int sq_n = 0;
+ uint32_t doorbell_hi;
+ uint64_t doorbell;
+ void *cq_db_reg = (char *)rxq->cq_uar + MLX5_CQ_DOORBELL;
+
+ sq_n = sq_n_rxq & MLX5_CQ_SQN_MASK;
+ doorbell_hi = sq_n << MLX5_CQ_SQN_OFFSET | (rxq->cq_ci & MLX5_CI_MASK);
+ doorbell = (uint64_t)doorbell_hi << 32;
+ doorbell |= rxq->cqn;
+ rxq->cq_db[MLX5_CQ_ARM_DB] = rte_cpu_to_be_32(doorbell_hi);
+ mlx5_uar_write64(rte_cpu_to_be_64(doorbell),
+ cq_db_reg, rxq->uar_lock_cq);
+}
+
+/**
+ * DPDK callback for Rx queue interrupt enable.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param rx_queue_id
+ * Rx queue number.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_rx_intr_enable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
+{
+ struct mlx5_rxq_ctrl *rxq_ctrl;
+
+ rxq_ctrl = mlx5_rxq_get(dev, rx_queue_id);
+ if (!rxq_ctrl)
+ goto error;
+ if (rxq_ctrl->irq) {
+ if (!rxq_ctrl->obj) {
+ mlx5_rxq_release(dev, rx_queue_id);
+ goto error;
+ }
+ mlx5_arm_cq(&rxq_ctrl->rxq, rxq_ctrl->rxq.cq_arm_sn);
+ }
+ mlx5_rxq_release(dev, rx_queue_id);
+ return 0;
+error:
+ rte_errno = EINVAL;
+ return -rte_errno;
+}
+
+/**
+ * DPDK callback for Rx queue interrupt disable.
+ *
+ * @param dev
+ * Pointer to Ethernet device structure.
+ * @param rx_queue_id
+ * Rx queue number.
+ *
+ * @return
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_rx_intr_disable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
struct mlx5_rxq_ctrl *rxq_ctrl;
int ret = 0;
- priv_lock(priv);
- rxq_data = (*priv->rxqs)[rx_queue_id];
- if (!rxq_data) {
- ret = EINVAL;
- goto exit;
+ rxq_ctrl = mlx5_rxq_get(dev, rx_queue_id);
+ if (!rxq_ctrl) {
+ rte_errno = EINVAL;
+ return -rte_errno;
+ }
+ if (!rxq_ctrl->obj)
+ goto error;
+ if (rxq_ctrl->irq) {
+ ret = priv->obj_ops.rxq_event_get(rxq_ctrl->obj);
+ if (ret < 0)
+ goto error;
+ rxq_ctrl->rxq.cq_arm_sn++;
+ }
+ mlx5_rxq_release(dev, rx_queue_id);
+ return 0;
+error:
+ /**
+ * The ret variable may be EAGAIN which means the get_event function was
+ * called before receiving one.
+ */
+ if (ret < 0)
+ rte_errno = errno;
+ else
+ rte_errno = EINVAL;
+ ret = rte_errno; /* Save rte_errno before cleanup. */
+ mlx5_rxq_release(dev, rx_queue_id);
+ if (ret != EAGAIN)
+ DRV_LOG(WARNING, "port %u unable to disable interrupt on Rx queue %d",
+ dev->data->port_id, rx_queue_id);
+ rte_errno = ret; /* Restore rte_errno. */
+ return -rte_errno;
+}
+
+/**
+ * Verify the Rx queue objects list is empty
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * The number of objects not released.
+ */
+int
+mlx5_rxq_obj_verify(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ int ret = 0;
+ struct mlx5_rxq_obj *rxq_obj;
+
+ LIST_FOREACH(rxq_obj, &priv->rxqsobj, next) {
+ DRV_LOG(DEBUG, "port %u Rx queue %u still referenced",
+ dev->data->port_id, rxq_obj->rxq_ctrl->rxq.idx);
+ ++ret;
}
- rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
- if (rxq_ctrl->irq) {
- struct mlx5_rxq_ibv *rxq_ibv;
+ return ret;
+}
- rxq_ibv = mlx5_priv_rxq_ibv_get(priv, rx_queue_id);
- if (!rxq_ibv) {
- ret = EINVAL;
- goto exit;
- }
- mlx5_arm_cq(rxq_data, rxq_data->cq_arm_sn);
- mlx5_priv_rxq_ibv_release(priv, rxq_ibv);
+/**
+ * Callback function to initialize mbufs for Multi-Packet RQ.
+ */
+static inline void
+mlx5_mprq_buf_init(struct rte_mempool *mp, void *opaque_arg,
+ void *_m, unsigned int i __rte_unused)
+{
+ struct mlx5_mprq_buf *buf = _m;
+ struct rte_mbuf_ext_shared_info *shinfo;
+ unsigned int strd_n = (unsigned int)(uintptr_t)opaque_arg;
+ unsigned int j;
+
+ memset(_m, 0, sizeof(*buf));
+ buf->mp = mp;
+ __atomic_store_n(&buf->refcnt, 1, __ATOMIC_RELAXED);
+ for (j = 0; j != strd_n; ++j) {
+ shinfo = &buf->shinfos[j];
+ shinfo->free_cb = mlx5_mprq_buf_free_cb;
+ shinfo->fcb_opaque = buf;
}
-exit:
- priv_unlock(priv);
- if (ret)
- WARN("unable to arm interrupt on rx queue %d", rx_queue_id);
- return -ret;
}
/**
- * DPDK callback for Rx queue interrupt disable.
+ * Free mempool of Multi-Packet RQ.
*
* @param dev
- * Pointer to Ethernet device structure.
- * @param rx_queue_id
- * Rx queue number.
+ * Pointer to Ethernet device.
*
* @return
- * 0 on success, negative on failure.
+ * 0 on success, negative errno value on failure.
*/
int
-mlx5_rx_intr_disable(struct rte_eth_dev *dev, uint16_t rx_queue_id)
+mlx5_mprq_free_mp(struct rte_eth_dev *dev)
{
- struct priv *priv = mlx5_get_priv(dev);
- struct mlx5_rxq_data *rxq_data;
- struct mlx5_rxq_ctrl *rxq_ctrl;
- struct mlx5_rxq_ibv *rxq_ibv = NULL;
- struct ibv_cq *ev_cq;
- void *ev_ctx;
- int ret = 0;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct rte_mempool *mp = priv->mprq_mp;
+ unsigned int i;
- priv_lock(priv);
- rxq_data = (*priv->rxqs)[rx_queue_id];
- if (!rxq_data) {
- ret = EINVAL;
- goto exit;
+ if (mp == NULL)
+ return 0;
+ DRV_LOG(DEBUG, "port %u freeing mempool (%s) for Multi-Packet RQ",
+ dev->data->port_id, mp->name);
+ /*
+ * If a buffer in the pool has been externally attached to a mbuf and it
+ * is still in use by application, destroying the Rx queue can spoil
+ * the packet. It is unlikely to happen but if application dynamically
+ * creates and destroys with holding Rx packets, this can happen.
+ *
+ * TODO: It is unavoidable for now because the mempool for Multi-Packet
+ * RQ isn't provided by application but managed by PMD.
+ */
+ if (!rte_mempool_full(mp)) {
+ DRV_LOG(ERR,
+ "port %u mempool for Multi-Packet RQ is still in use",
+ dev->data->port_id);
+ rte_errno = EBUSY;
+ return -rte_errno;
}
- rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
- if (!rxq_ctrl->irq)
- goto exit;
- rxq_ibv = mlx5_priv_rxq_ibv_get(priv, rx_queue_id);
- if (!rxq_ibv) {
- ret = EINVAL;
- goto exit;
+ rte_mempool_free(mp);
+ /* Unset mempool for each Rx queue. */
+ for (i = 0; i != priv->rxqs_n; ++i) {
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[i];
+
+ if (rxq == NULL)
+ continue;
+ rxq->mprq_mp = NULL;
+ }
+ priv->mprq_mp = NULL;
+ return 0;
+}
+
+/**
+ * Allocate a mempool for Multi-Packet RQ. All configured Rx queues share the
+ * mempool. If already allocated, reuse it if there're enough elements.
+ * Otherwise, resize it.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * 0 on success, negative errno value on failure.
+ */
+int
+mlx5_mprq_alloc_mp(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct rte_mempool *mp = priv->mprq_mp;
+ char name[RTE_MEMPOOL_NAMESIZE];
+ unsigned int desc = 0;
+ unsigned int buf_len;
+ unsigned int obj_num;
+ unsigned int obj_size;
+ unsigned int strd_num_n = 0;
+ unsigned int strd_sz_n = 0;
+ unsigned int i;
+ unsigned int n_ibv = 0;
+
+ if (!mlx5_mprq_enabled(dev))
+ return 0;
+ /* Count the total number of descriptors configured. */
+ for (i = 0; i != priv->rxqs_n; ++i) {
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[i];
+ struct mlx5_rxq_ctrl *rxq_ctrl = container_of
+ (rxq, struct mlx5_rxq_ctrl, rxq);
+
+ if (rxq == NULL || rxq_ctrl->type != MLX5_RXQ_TYPE_STANDARD)
+ continue;
+ n_ibv++;
+ desc += 1 << rxq->elts_n;
+ /* Get the max number of strides. */
+ if (strd_num_n < rxq->strd_num_n)
+ strd_num_n = rxq->strd_num_n;
+ /* Get the max size of a stride. */
+ if (strd_sz_n < rxq->strd_sz_n)
+ strd_sz_n = rxq->strd_sz_n;
}
- ret = ibv_get_cq_event(rxq_ibv->channel, &ev_cq, &ev_ctx);
- if (ret || ev_cq != rxq_ibv->cq) {
- ret = EINVAL;
+ MLX5_ASSERT(strd_num_n && strd_sz_n);
+ buf_len = (1 << strd_num_n) * (1 << strd_sz_n);
+ obj_size = sizeof(struct mlx5_mprq_buf) + buf_len + (1 << strd_num_n) *
+ sizeof(struct rte_mbuf_ext_shared_info) + RTE_PKTMBUF_HEADROOM;
+ /*
+ * Received packets can be either memcpy'd or externally referenced. In
+ * case that the packet is attached to an mbuf as an external buffer, as
+ * it isn't possible to predict how the buffers will be queued by
+ * application, there's no option to exactly pre-allocate needed buffers
+ * in advance but to speculatively prepares enough buffers.
+ *
+ * In the data path, if this Mempool is depleted, PMD will try to memcpy
+ * received packets to buffers provided by application (rxq->mp) until
+ * this Mempool gets available again.
+ */
+ desc *= 4;
+ obj_num = desc + MLX5_MPRQ_MP_CACHE_SZ * n_ibv;
+ /*
+ * rte_mempool_create_empty() has sanity check to refuse large cache
+ * size compared to the number of elements.
+ * CACHE_FLUSHTHRESH_MULTIPLIER is defined in a C file, so using a
+ * constant number 2 instead.
+ */
+ obj_num = RTE_MAX(obj_num, MLX5_MPRQ_MP_CACHE_SZ * 2);
+ /* Check a mempool is already allocated and if it can be resued. */
+ if (mp != NULL && mp->elt_size >= obj_size && mp->size >= obj_num) {
+ DRV_LOG(DEBUG, "port %u mempool %s is being reused",
+ dev->data->port_id, mp->name);
+ /* Reuse. */
goto exit;
+ } else if (mp != NULL) {
+ DRV_LOG(DEBUG, "port %u mempool %s should be resized, freeing it",
+ dev->data->port_id, mp->name);
+ /*
+ * If failed to free, which means it may be still in use, no way
+ * but to keep using the existing one. On buffer underrun,
+ * packets will be memcpy'd instead of external buffer
+ * attachment.
+ */
+ if (mlx5_mprq_free_mp(dev)) {
+ if (mp->elt_size >= obj_size)
+ goto exit;
+ else
+ return -rte_errno;
+ }
+ }
+ snprintf(name, sizeof(name), "port-%u-mprq", dev->data->port_id);
+ mp = rte_mempool_create(name, obj_num, obj_size, MLX5_MPRQ_MP_CACHE_SZ,
+ 0, NULL, NULL, mlx5_mprq_buf_init,
+ (void *)(uintptr_t)(1 << strd_num_n),
+ dev->device->numa_node, 0);
+ if (mp == NULL) {
+ DRV_LOG(ERR,
+ "port %u failed to allocate a mempool for"
+ " Multi-Packet RQ, count=%u, size=%u",
+ dev->data->port_id, obj_num, obj_size);
+ rte_errno = ENOMEM;
+ return -rte_errno;
}
- rxq_data->cq_arm_sn++;
- ibv_ack_cq_events(rxq_ibv->cq, 1);
+ priv->mprq_mp = mp;
exit:
- if (rxq_ibv)
- mlx5_priv_rxq_ibv_release(priv, rxq_ibv);
- priv_unlock(priv);
- if (ret)
- WARN("unable to disable interrupt on rx queue %d",
- rx_queue_id);
- return -ret;
+ /* Set mempool for each Rx queue. */
+ for (i = 0; i != priv->rxqs_n; ++i) {
+ struct mlx5_rxq_data *rxq = (*priv->rxqs)[i];
+ struct mlx5_rxq_ctrl *rxq_ctrl = container_of
+ (rxq, struct mlx5_rxq_ctrl, rxq);
+
+ if (rxq == NULL || rxq_ctrl->type != MLX5_RXQ_TYPE_STANDARD)
+ continue;
+ rxq->mprq_mp = mp;
+ }
+ DRV_LOG(INFO, "port %u Multi-Packet RQ is configured",
+ dev->data->port_id);
+ return 0;
+}
+
+#define MLX5_MAX_TCP_HDR_OFFSET ((unsigned int)(sizeof(struct rte_ether_hdr) + \
+ sizeof(struct rte_vlan_hdr) * 2 + \
+ sizeof(struct rte_ipv6_hdr)))
+#define MAX_TCP_OPTION_SIZE 40u
+#define MLX5_MAX_LRO_HEADER_FIX ((unsigned int)(MLX5_MAX_TCP_HDR_OFFSET + \
+ sizeof(struct rte_tcp_hdr) + \
+ MAX_TCP_OPTION_SIZE))
+
+/**
+ * Adjust the maximum LRO massage size.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * RX queue index.
+ * @param max_lro_size
+ * The maximum size for LRO packet.
+ */
+static void
+mlx5_max_lro_msg_size_adjust(struct rte_eth_dev *dev, uint16_t idx,
+ uint32_t max_lro_size)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+
+ if (priv->config.hca_attr.lro_max_msg_sz_mode ==
+ MLX5_LRO_MAX_MSG_SIZE_START_FROM_L4 && max_lro_size >
+ MLX5_MAX_TCP_HDR_OFFSET)
+ max_lro_size -= MLX5_MAX_TCP_HDR_OFFSET;
+ max_lro_size = RTE_MIN(max_lro_size, MLX5_MAX_LRO_SIZE);
+ MLX5_ASSERT(max_lro_size >= MLX5_LRO_SEG_CHUNK_SIZE);
+ max_lro_size /= MLX5_LRO_SEG_CHUNK_SIZE;
+ if (priv->max_lro_msg_size)
+ priv->max_lro_msg_size =
+ RTE_MIN((uint32_t)priv->max_lro_msg_size, max_lro_size);
+ else
+ priv->max_lro_msg_size = max_lro_size;
+ DRV_LOG(DEBUG,
+ "port %u Rx Queue %u max LRO message size adjusted to %u bytes",
+ dev->data->port_id, idx,
+ priv->max_lro_msg_size * MLX5_LRO_SEG_CHUNK_SIZE);
}
/**
- * Create the Rx queue Verbs object.
+ * Create a DPDK Rx queue.
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
* @param idx
- * Queue index in DPDK Rx queue array
+ * RX queue index.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param socket
+ * NUMA socket on which memory must be allocated.
*
* @return
- * The Verbs object initialised if it can be created.
+ * A DPDK queue object on success, NULL otherwise and rte_errno is set.
*/
-struct mlx5_rxq_ibv*
-mlx5_priv_rxq_ibv_new(struct priv *priv, uint16_t idx)
+struct mlx5_rxq_ctrl *
+mlx5_rxq_new(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
+ unsigned int socket, const struct rte_eth_rxconf *conf,
+ const struct rte_eth_rxseg_split *rx_seg, uint16_t n_seg)
{
- struct mlx5_rxq_data *rxq_data = (*priv->rxqs)[idx];
- struct mlx5_rxq_ctrl *rxq_ctrl =
- container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
- struct ibv_wq_attr mod;
- union {
- struct ibv_cq_init_attr_ex cq;
- struct ibv_wq_init_attr wq;
- struct ibv_cq_ex cq_attr;
- } attr;
- unsigned int cqe_n = (1 << rxq_data->elts_n) - 1;
- struct mlx5_rxq_ibv *tmpl;
- struct mlx5dv_cq cq_info;
- struct mlx5dv_rwq rwq;
- unsigned int i;
- int ret = 0;
- struct mlx5dv_obj obj;
-
- assert(rxq_data);
- assert(!rxq_ctrl->ibv);
- tmpl = rte_calloc_socket(__func__, 1, sizeof(*tmpl), 0,
- rxq_ctrl->socket);
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *tmpl;
+ unsigned int mb_len = rte_pktmbuf_data_room_size(rx_seg[0].mp);
+ struct mlx5_dev_config *config = &priv->config;
+ uint64_t offloads = conf->offloads |
+ dev->data->dev_conf.rxmode.offloads;
+ unsigned int lro_on_queue = !!(offloads & DEV_RX_OFFLOAD_TCP_LRO);
+ unsigned int max_rx_pkt_len = lro_on_queue ?
+ dev->data->dev_conf.rxmode.max_lro_pkt_size :
+ dev->data->dev_conf.rxmode.max_rx_pkt_len;
+ unsigned int non_scatter_min_mbuf_size = max_rx_pkt_len +
+ RTE_PKTMBUF_HEADROOM;
+ unsigned int max_lro_size = 0;
+ unsigned int first_mb_free_size = mb_len - RTE_PKTMBUF_HEADROOM;
+ const int mprq_en = mlx5_check_mprq_support(dev) > 0 && n_seg == 1 &&
+ !rx_seg[0].offset && !rx_seg[0].length;
+ unsigned int mprq_stride_nums = config->mprq.stride_num_n ?
+ config->mprq.stride_num_n : MLX5_MPRQ_STRIDE_NUM_N;
+ unsigned int mprq_stride_size = non_scatter_min_mbuf_size <=
+ (1U << config->mprq.max_stride_size_n) ?
+ log2above(non_scatter_min_mbuf_size) : MLX5_MPRQ_STRIDE_SIZE_N;
+ unsigned int mprq_stride_cap = (config->mprq.stride_num_n ?
+ (1U << config->mprq.stride_num_n) : (1U << mprq_stride_nums)) *
+ (config->mprq.stride_size_n ?
+ (1U << config->mprq.stride_size_n) : (1U << mprq_stride_size));
+ /*
+ * Always allocate extra slots, even if eventually
+ * the vector Rx will not be used.
+ */
+ uint16_t desc_n = desc + config->rx_vec_en * MLX5_VPMD_DESCS_PER_LOOP;
+ const struct rte_eth_rxseg_split *qs_seg = rx_seg;
+ unsigned int tail_len;
+
+ tmpl = mlx5_malloc(MLX5_MEM_RTE | MLX5_MEM_ZERO,
+ sizeof(*tmpl) + desc_n * sizeof(struct rte_mbuf *) +
+ (!!mprq_en) *
+ (desc >> mprq_stride_nums) * sizeof(struct mlx5_mprq_buf *),
+ 0, socket);
if (!tmpl) {
- ERROR("%p: cannot allocate verbs resources",
- (void *)rxq_ctrl);
- goto error;
+ rte_errno = ENOMEM;
+ return NULL;
}
- tmpl->rxq_ctrl = rxq_ctrl;
- /* Use the entire RX mempool as the memory region. */
- tmpl->mr = priv_mr_get(priv, rxq_data->mp);
- if (!tmpl->mr) {
- tmpl->mr = priv_mr_new(priv, rxq_data->mp);
- if (!tmpl->mr) {
- ERROR("%p: MR creation failure", (void *)rxq_ctrl);
+ MLX5_ASSERT(n_seg && n_seg <= MLX5_MAX_RXQ_NSEG);
+ /*
+ * Build the array of actual buffer offsets and lengths.
+ * Pad with the buffers from the last memory pool if
+ * needed to handle max size packets, replace zero length
+ * with the buffer length from the pool.
+ */
+ tail_len = max_rx_pkt_len;
+ do {
+ struct mlx5_eth_rxseg *hw_seg =
+ &tmpl->rxq.rxseg[tmpl->rxq.rxseg_n];
+ uint32_t buf_len, offset, seg_len;
+
+ /*
+ * For the buffers beyond descriptions offset is zero,
+ * the first buffer contains head room.
+ */
+ buf_len = rte_pktmbuf_data_room_size(qs_seg->mp);
+ offset = (tmpl->rxq.rxseg_n >= n_seg ? 0 : qs_seg->offset) +
+ (tmpl->rxq.rxseg_n ? 0 : RTE_PKTMBUF_HEADROOM);
+ /*
+ * For the buffers beyond descriptions the length is
+ * pool buffer length, zero lengths are replaced with
+ * pool buffer length either.
+ */
+ seg_len = tmpl->rxq.rxseg_n >= n_seg ? buf_len :
+ qs_seg->length ?
+ qs_seg->length :
+ (buf_len - offset);
+ /* Check is done in long int, now overflows. */
+ if (buf_len < seg_len + offset) {
+ DRV_LOG(ERR, "port %u Rx queue %u: Split offset/length "
+ "%u/%u can't be satisfied",
+ dev->data->port_id, idx,
+ qs_seg->length, qs_seg->offset);
+ rte_errno = EINVAL;
goto error;
}
- }
- if (rxq_ctrl->irq) {
- tmpl->channel = ibv_create_comp_channel(priv->ctx);
- if (!tmpl->channel) {
- ERROR("%p: Comp Channel creation failure",
- (void *)rxq_ctrl);
+ if (seg_len > tail_len)
+ seg_len = buf_len - offset;
+ if (++tmpl->rxq.rxseg_n > MLX5_MAX_RXQ_NSEG) {
+ DRV_LOG(ERR,
+ "port %u too many SGEs (%u) needed to handle"
+ " requested maximum packet size %u, the maximum"
+ " supported are %u", dev->data->port_id,
+ tmpl->rxq.rxseg_n, max_rx_pkt_len,
+ MLX5_MAX_RXQ_NSEG);
+ rte_errno = ENOTSUP;
goto error;
}
- }
- attr.cq = (struct ibv_cq_init_attr_ex){
- .comp_mask = 0,
- };
- if (priv->cqe_comp) {
- attr.cq.comp_mask |= IBV_CQ_INIT_ATTR_MASK_FLAGS;
- attr.cq.flags |= MLX5DV_CQ_INIT_ATTR_MASK_COMPRESSED_CQE;
+ /* Build the actual scattering element in the queue object. */
+ hw_seg->mp = qs_seg->mp;
+ MLX5_ASSERT(offset <= UINT16_MAX);
+ MLX5_ASSERT(seg_len <= UINT16_MAX);
+ hw_seg->offset = (uint16_t)offset;
+ hw_seg->length = (uint16_t)seg_len;
/*
- * For vectorized Rx, it must not be doubled in order to
- * make cq_ci and rq_ci aligned.
+ * Advance the segment descriptor, the padding is the based
+ * on the attributes of the last descriptor.
*/
- if (rxq_check_vec_support(rxq_data) < 0)
- cqe_n *= 2;
- }
- tmpl->cq = ibv_create_cq(priv->ctx, cqe_n, NULL, tmpl->channel, 0);
- if (tmpl->cq == NULL) {
- ERROR("%p: CQ creation failure", (void *)rxq_ctrl);
+ if (tmpl->rxq.rxseg_n < n_seg)
+ qs_seg++;
+ tail_len -= RTE_MIN(tail_len, seg_len);
+ } while (tail_len || !rte_is_power_of_2(tmpl->rxq.rxseg_n));
+ MLX5_ASSERT(tmpl->rxq.rxseg_n &&
+ tmpl->rxq.rxseg_n <= MLX5_MAX_RXQ_NSEG);
+ if (tmpl->rxq.rxseg_n > 1 && !(offloads & DEV_RX_OFFLOAD_SCATTER)) {
+ DRV_LOG(ERR, "port %u Rx queue %u: Scatter offload is not"
+ " configured and no enough mbuf space(%u) to contain "
+ "the maximum RX packet length(%u) with head-room(%u)",
+ dev->data->port_id, idx, mb_len, max_rx_pkt_len,
+ RTE_PKTMBUF_HEADROOM);
+ rte_errno = ENOSPC;
goto error;
}
- DEBUG("priv->device_attr.max_qp_wr is %d",
- priv->device_attr.orig_attr.max_qp_wr);
- DEBUG("priv->device_attr.max_sge is %d",
- priv->device_attr.orig_attr.max_sge);
- attr.wq = (struct ibv_wq_init_attr){
- .wq_context = NULL, /* Could be useful in the future. */
- .wq_type = IBV_WQT_RQ,
- /* Max number of outstanding WRs. */
- .max_wr = (1 << rxq_data->elts_n) >> rxq_data->sges_n,
- /* Max number of scatter/gather elements in a WR. */
- .max_sge = 1 << rxq_data->sges_n,
- .pd = priv->pd,
- .cq = tmpl->cq,
- .comp_mask =
- IBV_WQ_FLAGS_CVLAN_STRIPPING |
- 0,
- .create_flags = (rxq_data->vlan_strip ?
- IBV_WQ_FLAGS_CVLAN_STRIPPING :
- 0),
- };
- /* By default, FCS (CRC) is stripped by hardware. */
- if (rxq_data->crc_present) {
- attr.wq.create_flags |= IBV_WQ_FLAGS_SCATTER_FCS;
- attr.wq.comp_mask |= IBV_WQ_INIT_ATTR_FLAGS;
- }
-#ifdef HAVE_IBV_WQ_FLAG_RX_END_PADDING
- if (priv->hw_padding) {
- attr.wq.create_flags |= IBV_WQ_FLAG_RX_END_PADDING;
- attr.wq.comp_mask |= IBV_WQ_INIT_ATTR_FLAGS;
- }
-#endif
- tmpl->wq = ibv_create_wq(priv->ctx, &attr.wq);
- if (tmpl->wq == NULL) {
- ERROR("%p: WQ creation failure", (void *)rxq_ctrl);
+ tmpl->type = MLX5_RXQ_TYPE_STANDARD;
+ if (mlx5_mr_btree_init(&tmpl->rxq.mr_ctrl.cache_bh,
+ MLX5_MR_BTREE_CACHE_N, socket)) {
+ /* rte_errno is already set. */
goto error;
}
+ tmpl->socket = socket;
+ if (dev->data->dev_conf.intr_conf.rxq)
+ tmpl->irq = 1;
/*
- * Make sure number of WRs*SGEs match expectations since a queue
- * cannot allocate more than "desc" buffers.
+ * This Rx queue can be configured as a Multi-Packet RQ if all of the
+ * following conditions are met:
+ * - MPRQ is enabled.
+ * - The number of descs is more than the number of strides.
+ * - max_rx_pkt_len plus overhead is less than the max size
+ * of a stride or mprq_stride_size is specified by a user.
+ * Need to make sure that there are enough strides to encap
+ * the maximum packet size in case mprq_stride_size is set.
+ * Otherwise, enable Rx scatter if necessary.
*/
- if (((int)attr.wq.max_wr !=
- ((1 << rxq_data->elts_n) >> rxq_data->sges_n)) ||
- ((int)attr.wq.max_sge != (1 << rxq_data->sges_n))) {
- ERROR("%p: requested %u*%u but got %u*%u WRs*SGEs",
- (void *)rxq_ctrl,
- ((1 << rxq_data->elts_n) >> rxq_data->sges_n),
- (1 << rxq_data->sges_n),
- attr.wq.max_wr, attr.wq.max_sge);
- goto error;
+ if (mprq_en && desc > (1U << mprq_stride_nums) &&
+ (non_scatter_min_mbuf_size <=
+ (1U << config->mprq.max_stride_size_n) ||
+ (config->mprq.stride_size_n &&
+ non_scatter_min_mbuf_size <= mprq_stride_cap))) {
+ /* TODO: Rx scatter isn't supported yet. */
+ tmpl->rxq.sges_n = 0;
+ /* Trim the number of descs needed. */
+ desc >>= mprq_stride_nums;
+ tmpl->rxq.strd_num_n = config->mprq.stride_num_n ?
+ config->mprq.stride_num_n : mprq_stride_nums;
+ tmpl->rxq.strd_sz_n = config->mprq.stride_size_n ?
+ config->mprq.stride_size_n : mprq_stride_size;
+ tmpl->rxq.strd_shift_en = MLX5_MPRQ_TWO_BYTE_SHIFT;
+ tmpl->rxq.strd_scatter_en =
+ !!(offloads & DEV_RX_OFFLOAD_SCATTER);
+ tmpl->rxq.mprq_max_memcpy_len = RTE_MIN(first_mb_free_size,
+ config->mprq.max_memcpy_len);
+ max_lro_size = RTE_MIN(max_rx_pkt_len,
+ (1u << tmpl->rxq.strd_num_n) *
+ (1u << tmpl->rxq.strd_sz_n));
+ DRV_LOG(DEBUG,
+ "port %u Rx queue %u: Multi-Packet RQ is enabled"
+ " strd_num_n = %u, strd_sz_n = %u",
+ dev->data->port_id, idx,
+ tmpl->rxq.strd_num_n, tmpl->rxq.strd_sz_n);
+ } else if (tmpl->rxq.rxseg_n == 1) {
+ MLX5_ASSERT(max_rx_pkt_len <= first_mb_free_size);
+ tmpl->rxq.sges_n = 0;
+ max_lro_size = max_rx_pkt_len;
+ } else if (offloads & DEV_RX_OFFLOAD_SCATTER) {
+ unsigned int sges_n;
+
+ if (lro_on_queue && first_mb_free_size <
+ MLX5_MAX_LRO_HEADER_FIX) {
+ DRV_LOG(ERR, "Not enough space in the first segment(%u)"
+ " to include the max header size(%u) for LRO",
+ first_mb_free_size, MLX5_MAX_LRO_HEADER_FIX);
+ rte_errno = ENOTSUP;
+ goto error;
+ }
+ /*
+ * Determine the number of SGEs needed for a full packet
+ * and round it to the next power of two.
+ */
+ sges_n = log2above(tmpl->rxq.rxseg_n);
+ if (sges_n > MLX5_MAX_LOG_RQ_SEGS) {
+ DRV_LOG(ERR,
+ "port %u too many SGEs (%u) needed to handle"
+ " requested maximum packet size %u, the maximum"
+ " supported are %u", dev->data->port_id,
+ 1 << sges_n, max_rx_pkt_len,
+ 1u << MLX5_MAX_LOG_RQ_SEGS);
+ rte_errno = ENOTSUP;
+ goto error;
+ }
+ tmpl->rxq.sges_n = sges_n;
+ max_lro_size = max_rx_pkt_len;
}
- /* Change queue state to ready. */
- mod = (struct ibv_wq_attr){
- .attr_mask = IBV_WQ_ATTR_STATE,
- .wq_state = IBV_WQS_RDY,
- };
- ret = ibv_modify_wq(tmpl->wq, &mod);
- if (ret) {
- ERROR("%p: WQ state to IBV_WQS_RDY failed",
- (void *)rxq_ctrl);
+ if (config->mprq.enabled && !mlx5_rxq_mprq_enabled(&tmpl->rxq))
+ DRV_LOG(WARNING,
+ "port %u MPRQ is requested but cannot be enabled\n"
+ " (requested: pkt_sz = %u, desc_num = %u,"
+ " rxq_num = %u, stride_sz = %u, stride_num = %u\n"
+ " supported: min_rxqs_num = %u,"
+ " min_stride_sz = %u, max_stride_sz = %u).",
+ dev->data->port_id, non_scatter_min_mbuf_size,
+ desc, priv->rxqs_n,
+ config->mprq.stride_size_n ?
+ (1U << config->mprq.stride_size_n) :
+ (1U << mprq_stride_size),
+ config->mprq.stride_num_n ?
+ (1U << config->mprq.stride_num_n) :
+ (1U << mprq_stride_nums),
+ config->mprq.min_rxqs_num,
+ (1U << config->mprq.min_stride_size_n),
+ (1U << config->mprq.max_stride_size_n));
+ DRV_LOG(DEBUG, "port %u maximum number of segments per packet: %u",
+ dev->data->port_id, 1 << tmpl->rxq.sges_n);
+ if (desc % (1 << tmpl->rxq.sges_n)) {
+ DRV_LOG(ERR,
+ "port %u number of Rx queue descriptors (%u) is not a"
+ " multiple of SGEs per packet (%u)",
+ dev->data->port_id,
+ desc,
+ 1 << tmpl->rxq.sges_n);
+ rte_errno = EINVAL;
goto error;
}
- obj.cq.in = tmpl->cq;
- obj.cq.out = &cq_info;
- obj.rwq.in = tmpl->wq;
- obj.rwq.out = &rwq;
- ret = mlx5dv_init_obj(&obj, MLX5DV_OBJ_CQ | MLX5DV_OBJ_RWQ);
- if (ret != 0)
- goto error;
- if (cq_info.cqe_size != RTE_CACHE_LINE_SIZE) {
- ERROR("Wrong MLX5_CQE_SIZE environment variable value: "
- "it should be set to %u", RTE_CACHE_LINE_SIZE);
+ mlx5_max_lro_msg_size_adjust(dev, idx, max_lro_size);
+ /* Toggle RX checksum offload if hardware supports it. */
+ tmpl->rxq.csum = !!(offloads & DEV_RX_OFFLOAD_CHECKSUM);
+ /* Configure Rx timestamp. */
+ tmpl->rxq.hw_timestamp = !!(offloads & DEV_RX_OFFLOAD_TIMESTAMP);
+ tmpl->rxq.timestamp_rx_flag = 0;
+ if (tmpl->rxq.hw_timestamp && rte_mbuf_dyn_rx_timestamp_register(
+ &tmpl->rxq.timestamp_offset,
+ &tmpl->rxq.timestamp_rx_flag) != 0) {
+ DRV_LOG(ERR, "Cannot register Rx timestamp field/flag");
goto error;
}
- /* Fill the rings. */
- rxq_data->wqes = (volatile struct mlx5_wqe_data_seg (*)[])
- (uintptr_t)rwq.buf;
- for (i = 0; (i != (unsigned int)(1 << rxq_data->elts_n)); ++i) {
- struct rte_mbuf *buf = (*rxq_data->elts)[i];
- volatile struct mlx5_wqe_data_seg *scat = &(*rxq_data->wqes)[i];
-
- /* scat->addr must be able to store a pointer. */
- assert(sizeof(scat->addr) >= sizeof(uintptr_t));
- *scat = (struct mlx5_wqe_data_seg){
- .addr = rte_cpu_to_be_64(rte_pktmbuf_mtod(buf,
- uintptr_t)),
- .byte_count = rte_cpu_to_be_32(DATA_LEN(buf)),
- .lkey = tmpl->mr->lkey,
- };
- }
- rxq_data->rq_db = rwq.dbrec;
- rxq_data->cqe_n = log2above(cq_info.cqe_cnt);
- rxq_data->cq_ci = 0;
- rxq_data->rq_ci = 0;
- rxq_data->rq_pi = 0;
- rxq_data->zip = (struct rxq_zip){
- .ai = 0,
- };
- rxq_data->cq_db = cq_info.dbrec;
- rxq_data->cqes = (volatile struct mlx5_cqe (*)[])(uintptr_t)cq_info.buf;
- /* Update doorbell counter. */
- rxq_data->rq_ci = (1 << rxq_data->elts_n) >> rxq_data->sges_n;
- rte_wmb();
- *rxq_data->rq_db = rte_cpu_to_be_32(rxq_data->rq_ci);
- DEBUG("%p: rxq updated with %p", (void *)rxq_ctrl, (void *)&tmpl);
- rte_atomic32_inc(&tmpl->refcnt);
- DEBUG("%p: Verbs Rx queue %p: refcnt %d", (void *)priv,
- (void *)tmpl, rte_atomic32_read(&tmpl->refcnt));
- LIST_INSERT_HEAD(&priv->rxqsibv, tmpl, next);
+ /* Configure VLAN stripping. */
+ tmpl->rxq.vlan_strip = !!(offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
+ /* By default, FCS (CRC) is stripped by hardware. */
+ tmpl->rxq.crc_present = 0;
+ tmpl->rxq.lro = lro_on_queue;
+ if (offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
+ if (config->hw_fcs_strip) {
+ /*
+ * RQs used for LRO-enabled TIRs should not be
+ * configured to scatter the FCS.
+ */
+ if (lro_on_queue)
+ DRV_LOG(WARNING,
+ "port %u CRC stripping has been "
+ "disabled but will still be performed "
+ "by hardware, because LRO is enabled",
+ dev->data->port_id);
+ else
+ tmpl->rxq.crc_present = 1;
+ } else {
+ DRV_LOG(WARNING,
+ "port %u CRC stripping has been disabled but will"
+ " still be performed by hardware, make sure MLNX_OFED"
+ " and firmware are up to date",
+ dev->data->port_id);
+ }
+ }
+ DRV_LOG(DEBUG,
+ "port %u CRC stripping is %s, %u bytes will be subtracted from"
+ " incoming frames to hide it",
+ dev->data->port_id,
+ tmpl->rxq.crc_present ? "disabled" : "enabled",
+ tmpl->rxq.crc_present << 2);
+ /* Save port ID. */
+ tmpl->rxq.rss_hash = !!priv->rss_conf.rss_hf &&
+ (!!(dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS));
+ tmpl->rxq.port_id = dev->data->port_id;
+ tmpl->priv = priv;
+ tmpl->rxq.mp = rx_seg[0].mp;
+ tmpl->rxq.elts_n = log2above(desc);
+ tmpl->rxq.rq_repl_thresh =
+ MLX5_VPMD_RXQ_RPLNSH_THRESH(desc_n);
+ tmpl->rxq.elts =
+ (struct rte_mbuf *(*)[desc_n])(tmpl + 1);
+ tmpl->rxq.mprq_bufs =
+ (struct mlx5_mprq_buf *(*)[desc])(*tmpl->rxq.elts + desc_n);
+#ifndef RTE_ARCH_64
+ tmpl->rxq.uar_lock_cq = &priv->sh->uar_lock_cq;
+#endif
+ tmpl->rxq.idx = idx;
+ __atomic_fetch_add(&tmpl->refcnt, 1, __ATOMIC_RELAXED);
+ LIST_INSERT_HEAD(&priv->rxqsctrl, tmpl, next);
return tmpl;
error:
- if (tmpl->wq)
- claim_zero(ibv_destroy_wq(tmpl->wq));
- if (tmpl->cq)
- claim_zero(ibv_destroy_cq(tmpl->cq));
- if (tmpl->channel)
- claim_zero(ibv_destroy_comp_channel(tmpl->channel));
- if (tmpl->mr)
- priv_mr_release(priv, tmpl->mr);
+ mlx5_mr_btree_free(&tmpl->rxq.mr_ctrl.cache_bh);
+ mlx5_free(tmpl);
return NULL;
}
/**
- * Get an Rx queue Verbs object.
+ * Create a DPDK Rx hairpin queue.
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
* @param idx
- * Queue index in DPDK Rx queue array
+ * RX queue index.
+ * @param desc
+ * Number of descriptors to configure in queue.
+ * @param hairpin_conf
+ * The hairpin binding configuration.
*
* @return
- * The Verbs object if it exists.
+ * A DPDK queue object on success, NULL otherwise and rte_errno is set.
*/
-struct mlx5_rxq_ibv*
-mlx5_priv_rxq_ibv_get(struct priv *priv, uint16_t idx)
+struct mlx5_rxq_ctrl *
+mlx5_rxq_hairpin_new(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc,
+ const struct rte_eth_hairpin_conf *hairpin_conf)
{
- struct mlx5_rxq_data *rxq_data = (*priv->rxqs)[idx];
- struct mlx5_rxq_ctrl *rxq_ctrl;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *tmpl;
- if (idx >= priv->rxqs_n)
- return NULL;
- if (!rxq_data)
+ tmpl = mlx5_malloc(MLX5_MEM_RTE | MLX5_MEM_ZERO, sizeof(*tmpl), 0,
+ SOCKET_ID_ANY);
+ if (!tmpl) {
+ rte_errno = ENOMEM;
return NULL;
- rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
- if (rxq_ctrl->ibv) {
- priv_mr_get(priv, rxq_data->mp);
- rte_atomic32_inc(&rxq_ctrl->ibv->refcnt);
- DEBUG("%p: Verbs Rx queue %p: refcnt %d", (void *)priv,
- (void *)rxq_ctrl->ibv,
- rte_atomic32_read(&rxq_ctrl->ibv->refcnt));
- }
- return rxq_ctrl->ibv;
+ }
+ tmpl->type = MLX5_RXQ_TYPE_HAIRPIN;
+ tmpl->socket = SOCKET_ID_ANY;
+ tmpl->rxq.rss_hash = 0;
+ tmpl->rxq.port_id = dev->data->port_id;
+ tmpl->priv = priv;
+ tmpl->rxq.mp = NULL;
+ tmpl->rxq.elts_n = log2above(desc);
+ tmpl->rxq.elts = NULL;
+ tmpl->rxq.mr_ctrl.cache_bh = (struct mlx5_mr_btree) { 0 };
+ tmpl->hairpin_conf = *hairpin_conf;
+ tmpl->rxq.idx = idx;
+ __atomic_fetch_add(&tmpl->refcnt, 1, __ATOMIC_RELAXED);
+ LIST_INSERT_HEAD(&priv->rxqsctrl, tmpl, next);
+ return tmpl;
}
/**
- * Release an Rx verbs queue object.
+ * Get a Rx queue.
*
- * @param priv
- * Pointer to private structure.
- * @param rxq_ibv
- * Verbs Rx queue object.
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * RX queue index.
*
* @return
- * 0 on success, errno value on failure.
+ * A pointer to the queue if it exists, NULL otherwise.
+ */
+struct mlx5_rxq_ctrl *
+mlx5_rxq_get(struct rte_eth_dev *dev, uint16_t idx)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_data *rxq_data = (*priv->rxqs)[idx];
+ struct mlx5_rxq_ctrl *rxq_ctrl = NULL;
+
+ if (rxq_data) {
+ rxq_ctrl = container_of(rxq_data, struct mlx5_rxq_ctrl, rxq);
+ __atomic_fetch_add(&rxq_ctrl->refcnt, 1, __ATOMIC_RELAXED);
+ }
+ return rxq_ctrl;
+}
+
+/**
+ * Release a Rx queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * RX queue index.
+ *
+ * @return
+ * 1 while a reference on it exists, 0 when freed.
*/
int
-mlx5_priv_rxq_ibv_release(struct priv *priv, struct mlx5_rxq_ibv *rxq_ibv)
+mlx5_rxq_release(struct rte_eth_dev *dev, uint16_t idx)
{
- int ret;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *rxq_ctrl;
- assert(rxq_ibv);
- assert(rxq_ibv->wq);
- assert(rxq_ibv->cq);
- assert(rxq_ibv->mr);
- ret = priv_mr_release(priv, rxq_ibv->mr);
- if (!ret)
- rxq_ibv->mr = NULL;
- DEBUG("%p: Verbs Rx queue %p: refcnt %d", (void *)priv,
- (void *)rxq_ibv, rte_atomic32_read(&rxq_ibv->refcnt));
- if (rte_atomic32_dec_and_test(&rxq_ibv->refcnt)) {
- rxq_free_elts(rxq_ibv->rxq_ctrl);
- claim_zero(ibv_destroy_wq(rxq_ibv->wq));
- claim_zero(ibv_destroy_cq(rxq_ibv->cq));
- if (rxq_ibv->channel)
- claim_zero(ibv_destroy_comp_channel(rxq_ibv->channel));
- LIST_REMOVE(rxq_ibv, next);
- rte_free(rxq_ibv);
+ if (!(*priv->rxqs)[idx])
return 0;
+ rxq_ctrl = container_of((*priv->rxqs)[idx], struct mlx5_rxq_ctrl, rxq);
+ if (__atomic_sub_fetch(&rxq_ctrl->refcnt, 1, __ATOMIC_RELAXED) > 1)
+ return 1;
+ if (rxq_ctrl->obj) {
+ priv->obj_ops.rxq_obj_release(rxq_ctrl->obj);
+ LIST_REMOVE(rxq_ctrl->obj, next);
+ mlx5_free(rxq_ctrl->obj);
+ rxq_ctrl->obj = NULL;
+ }
+ if (rxq_ctrl->type == MLX5_RXQ_TYPE_STANDARD) {
+ rxq_free_elts(rxq_ctrl);
+ dev->data->rx_queue_state[idx] = RTE_ETH_QUEUE_STATE_STOPPED;
}
- return EBUSY;
+ if (!__atomic_load_n(&rxq_ctrl->refcnt, __ATOMIC_RELAXED)) {
+ if (rxq_ctrl->type == MLX5_RXQ_TYPE_STANDARD)
+ mlx5_mr_btree_free(&rxq_ctrl->rxq.mr_ctrl.cache_bh);
+ LIST_REMOVE(rxq_ctrl, next);
+ mlx5_free(rxq_ctrl);
+ (*priv->rxqs)[idx] = NULL;
+ }
+ return 0;
}
/**
- * Verify the Verbs Rx queue list is empty
+ * Verify the Rx Queue list is empty
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
*
- * @return the number of object not released.
+ * @return
+ * The number of object not released.
*/
int
-mlx5_priv_rxq_ibv_verify(struct priv *priv)
+mlx5_rxq_verify(struct rte_eth_dev *dev)
{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *rxq_ctrl;
int ret = 0;
- struct mlx5_rxq_ibv *rxq_ibv;
- LIST_FOREACH(rxq_ibv, &priv->rxqsibv, next) {
- DEBUG("%p: Verbs Rx queue %p still referenced", (void *)priv,
- (void *)rxq_ibv);
+ LIST_FOREACH(rxq_ctrl, &priv->rxqsctrl, next) {
+ DRV_LOG(DEBUG, "port %u Rx Queue %u still referenced",
+ dev->data->port_id, rxq_ctrl->rxq.idx);
++ret;
}
return ret;
}
/**
- * Return true if a single reference exists on the object.
+ * Get a Rx queue type.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param idx
+ * Rx queue index.
*
- * @param priv
- * Pointer to private structure.
- * @param rxq_ibv
- * Verbs Rx queue object.
+ * @return
+ * The Rx queue type.
*/
-int
-mlx5_priv_rxq_ibv_releasable(struct priv *priv, struct mlx5_rxq_ibv *rxq_ibv)
+enum mlx5_rxq_type
+mlx5_rxq_get_type(struct rte_eth_dev *dev, uint16_t idx)
{
- (void)priv;
- assert(rxq_ibv);
- return (rte_atomic32_read(&rxq_ibv->refcnt) == 1);
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *rxq_ctrl = NULL;
+
+ if (idx < priv->rxqs_n && (*priv->rxqs)[idx]) {
+ rxq_ctrl = container_of((*priv->rxqs)[idx],
+ struct mlx5_rxq_ctrl,
+ rxq);
+ return rxq_ctrl->type;
+ }
+ return MLX5_RXQ_TYPE_UNDEFINED;
}
-/**
- * Create a DPDK Rx queue.
+/*
+ * Get a Rx hairpin queue configuration.
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
* @param idx
- * TX queue index.
- * @param desc
- * Number of descriptors to configure in queue.
- * @param socket
- * NUMA socket on which memory must be allocated.
+ * Rx queue index.
*
* @return
- * A DPDK queue object on success.
+ * Pointer to the configuration if a hairpin RX queue, otherwise NULL.
*/
-struct mlx5_rxq_ctrl*
-mlx5_priv_rxq_new(struct priv *priv, uint16_t idx, uint16_t desc,
- unsigned int socket, struct rte_mempool *mp)
+const struct rte_eth_hairpin_conf *
+mlx5_rxq_get_hairpin_conf(struct rte_eth_dev *dev, uint16_t idx)
{
- struct rte_eth_dev *dev = priv->dev;
- struct mlx5_rxq_ctrl *tmpl;
- const uint16_t desc_n =
- desc + priv->rx_vec_en * MLX5_VPMD_DESCS_PER_LOOP;
- unsigned int mb_len = rte_pktmbuf_data_room_size(mp);
-
- tmpl = rte_calloc_socket("RXQ", 1,
- sizeof(*tmpl) +
- desc_n * sizeof(struct rte_mbuf *),
- 0, socket);
- if (!tmpl)
- return NULL;
- if (priv->dev->data->dev_conf.intr_conf.rxq)
- tmpl->irq = 1;
- /* Enable scattered packets support for this queue if necessary. */
- assert(mb_len >= RTE_PKTMBUF_HEADROOM);
- if (dev->data->dev_conf.rxmode.max_rx_pkt_len <=
- (mb_len - RTE_PKTMBUF_HEADROOM)) {
- tmpl->rxq.sges_n = 0;
- } else if (dev->data->dev_conf.rxmode.enable_scatter) {
- unsigned int size =
- RTE_PKTMBUF_HEADROOM +
- dev->data->dev_conf.rxmode.max_rx_pkt_len;
- unsigned int sges_n;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_rxq_ctrl *rxq_ctrl = NULL;
- /*
- * Determine the number of SGEs needed for a full packet
- * and round it to the next power of two.
- */
- sges_n = log2above((size / mb_len) + !!(size % mb_len));
- tmpl->rxq.sges_n = sges_n;
- /* Make sure rxq.sges_n did not overflow. */
- size = mb_len * (1 << tmpl->rxq.sges_n);
- size -= RTE_PKTMBUF_HEADROOM;
- if (size < dev->data->dev_conf.rxmode.max_rx_pkt_len) {
- ERROR("%p: too many SGEs (%u) needed to handle"
- " requested maximum packet size %u",
- (void *)dev,
- 1 << sges_n,
- dev->data->dev_conf.rxmode.max_rx_pkt_len);
- goto error;
- }
- } else {
- WARN("%p: the requested maximum Rx packet size (%u) is"
- " larger than a single mbuf (%u) and scattered"
- " mode has not been requested",
- (void *)dev,
- dev->data->dev_conf.rxmode.max_rx_pkt_len,
- mb_len - RTE_PKTMBUF_HEADROOM);
- }
- DEBUG("%p: maximum number of segments per packet: %u",
- (void *)dev, 1 << tmpl->rxq.sges_n);
- if (desc % (1 << tmpl->rxq.sges_n)) {
- ERROR("%p: number of RX queue descriptors (%u) is not a"
- " multiple of SGEs per packet (%u)",
- (void *)dev,
- desc,
- 1 << tmpl->rxq.sges_n);
- goto error;
+ if (idx < priv->rxqs_n && (*priv->rxqs)[idx]) {
+ rxq_ctrl = container_of((*priv->rxqs)[idx],
+ struct mlx5_rxq_ctrl,
+ rxq);
+ if (rxq_ctrl->type == MLX5_RXQ_TYPE_HAIRPIN)
+ return &rxq_ctrl->hairpin_conf;
}
- /* Toggle RX checksum offload if hardware supports it. */
- if (priv->hw_csum)
- tmpl->rxq.csum = !!dev->data->dev_conf.rxmode.hw_ip_checksum;
- if (priv->hw_csum_l2tun)
- tmpl->rxq.csum_l2tun =
- !!dev->data->dev_conf.rxmode.hw_ip_checksum;
- /* Configure VLAN stripping. */
- tmpl->rxq.vlan_strip = (priv->hw_vlan_strip &&
- !!dev->data->dev_conf.rxmode.hw_vlan_strip);
- /* By default, FCS (CRC) is stripped by hardware. */
- if (dev->data->dev_conf.rxmode.hw_strip_crc) {
- tmpl->rxq.crc_present = 0;
- } else if (priv->hw_fcs_strip) {
- tmpl->rxq.crc_present = 1;
- } else {
- WARN("%p: CRC stripping has been disabled but will still"
- " be performed by hardware, make sure MLNX_OFED and"
- " firmware are up to date",
- (void *)dev);
- tmpl->rxq.crc_present = 0;
- }
- DEBUG("%p: CRC stripping is %s, %u bytes will be subtracted from"
- " incoming frames to hide it",
- (void *)dev,
- tmpl->rxq.crc_present ? "disabled" : "enabled",
- tmpl->rxq.crc_present << 2);
- /* Save port ID. */
- tmpl->rxq.rss_hash = priv->rxqs_n > 1;
- tmpl->rxq.port_id = dev->data->port_id;
- tmpl->priv = priv;
- tmpl->rxq.mp = mp;
- tmpl->rxq.stats.idx = idx;
- tmpl->rxq.elts_n = log2above(desc);
- tmpl->rxq.elts =
- (struct rte_mbuf *(*)[1 << tmpl->rxq.elts_n])(tmpl + 1);
- rte_atomic32_inc(&tmpl->refcnt);
- DEBUG("%p: Rx queue %p: refcnt %d", (void *)priv,
- (void *)tmpl, rte_atomic32_read(&tmpl->refcnt));
- LIST_INSERT_HEAD(&priv->rxqsctrl, tmpl, next);
- return tmpl;
-error:
- rte_free(tmpl);
return NULL;
}
/**
- * Get a Rx queue.
+ * Match queues listed in arguments to queues contained in indirection table
+ * object.
*
- * @param priv
- * Pointer to private structure.
- * @param idx
- * TX queue index.
+ * @param ind_tbl
+ * Pointer to indirection table to match.
+ * @param queues
+ * Queues to match to ques in indirection table.
+ * @param queues_n
+ * Number of queues in the array.
*
* @return
- * A pointer to the queue if it exists.
+ * 1 if all queues in indirection table match 0 othrwise.
*/
-struct mlx5_rxq_ctrl*
-mlx5_priv_rxq_get(struct priv *priv, uint16_t idx)
+static int
+mlx5_ind_table_obj_match_queues(const struct mlx5_ind_table_obj *ind_tbl,
+ const uint16_t *queues, uint32_t queues_n)
{
- struct mlx5_rxq_ctrl *rxq_ctrl = NULL;
-
- if ((*priv->rxqs)[idx]) {
- rxq_ctrl = container_of((*priv->rxqs)[idx],
- struct mlx5_rxq_ctrl,
- rxq);
-
- mlx5_priv_rxq_ibv_get(priv, idx);
- rte_atomic32_inc(&rxq_ctrl->refcnt);
- DEBUG("%p: Rx queue %p: refcnt %d", (void *)priv,
- (void *)rxq_ctrl, rte_atomic32_read(&rxq_ctrl->refcnt));
- }
- return rxq_ctrl;
+ return (ind_tbl->queues_n == queues_n) &&
+ (!memcmp(ind_tbl->queues, queues,
+ ind_tbl->queues_n * sizeof(ind_tbl->queues[0])));
}
/**
- * Release a Rx queue.
+ * Get an indirection table.
*
- * @param priv
- * Pointer to private structure.
- * @param idx
- * TX queue index.
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param queues
+ * Queues entering in the indirection table.
+ * @param queues_n
+ * Number of queues in the array.
*
* @return
- * 0 on success, errno value on failure.
+ * An indirection table if found.
*/
-int
-mlx5_priv_rxq_release(struct priv *priv, uint16_t idx)
+struct mlx5_ind_table_obj *
+mlx5_ind_table_obj_get(struct rte_eth_dev *dev, const uint16_t *queues,
+ uint32_t queues_n)
{
- struct mlx5_rxq_ctrl *rxq_ctrl;
-
- if (!(*priv->rxqs)[idx])
- return 0;
- rxq_ctrl = container_of((*priv->rxqs)[idx], struct mlx5_rxq_ctrl, rxq);
- assert(rxq_ctrl->priv);
- if (rxq_ctrl->ibv) {
- int ret;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl;
- ret = mlx5_priv_rxq_ibv_release(rxq_ctrl->priv, rxq_ctrl->ibv);
- if (!ret)
- rxq_ctrl->ibv = NULL;
+ LIST_FOREACH(ind_tbl, &priv->ind_tbls, next) {
+ if ((ind_tbl->queues_n == queues_n) &&
+ (memcmp(ind_tbl->queues, queues,
+ ind_tbl->queues_n * sizeof(ind_tbl->queues[0]))
+ == 0))
+ break;
}
- DEBUG("%p: Rx queue %p: refcnt %d", (void *)priv,
- (void *)rxq_ctrl, rte_atomic32_read(&rxq_ctrl->refcnt));
- if (rte_atomic32_dec_and_test(&rxq_ctrl->refcnt)) {
- LIST_REMOVE(rxq_ctrl, next);
- rte_free(rxq_ctrl);
- (*priv->rxqs)[idx] = NULL;
- return 0;
+ if (ind_tbl) {
+ unsigned int i;
+
+ __atomic_fetch_add(&ind_tbl->refcnt, 1, __ATOMIC_RELAXED);
+ for (i = 0; i != ind_tbl->queues_n; ++i)
+ mlx5_rxq_get(dev, ind_tbl->queues[i]);
}
- return EBUSY;
+ return ind_tbl;
}
/**
- * Verify if the queue can be released.
+ * Release an indirection table.
*
- * @param priv
- * Pointer to private structure.
- * @param idx
- * TX queue index.
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param ind_table
+ * Indirection table to release.
+ * @param standalone
+ * Indirection table for Standalone queue.
*
* @return
- * 1 if the queue can be released.
+ * 1 while a reference on it exists, 0 when freed.
*/
int
-mlx5_priv_rxq_releasable(struct priv *priv, uint16_t idx)
+mlx5_ind_table_obj_release(struct rte_eth_dev *dev,
+ struct mlx5_ind_table_obj *ind_tbl,
+ bool standalone)
{
- struct mlx5_rxq_ctrl *rxq_ctrl;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ unsigned int i;
- if (!(*priv->rxqs)[idx])
- return -1;
- rxq_ctrl = container_of((*priv->rxqs)[idx], struct mlx5_rxq_ctrl, rxq);
- return (rte_atomic32_read(&rxq_ctrl->refcnt) == 1);
+ if (__atomic_sub_fetch(&ind_tbl->refcnt, 1, __ATOMIC_RELAXED) == 0)
+ priv->obj_ops.ind_table_destroy(ind_tbl);
+ for (i = 0; i != ind_tbl->queues_n; ++i)
+ claim_nonzero(mlx5_rxq_release(dev, ind_tbl->queues[i]));
+ if (__atomic_load_n(&ind_tbl->refcnt, __ATOMIC_RELAXED) == 0) {
+ if (!standalone)
+ LIST_REMOVE(ind_tbl, next);
+ mlx5_free(ind_tbl);
+ return 0;
+ }
+ return 1;
}
/**
* Verify the Rx Queue list is empty
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
*
- * @return the number of object not released.
+ * @return
+ * The number of object not released.
*/
int
-mlx5_priv_rxq_verify(struct priv *priv)
+mlx5_ind_table_obj_verify(struct rte_eth_dev *dev)
{
- struct mlx5_rxq_ctrl *rxq_ctrl;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl;
int ret = 0;
- LIST_FOREACH(rxq_ctrl, &priv->rxqsctrl, next) {
- DEBUG("%p: Rx Queue %p still referenced", (void *)priv,
- (void *)rxq_ctrl);
+ LIST_FOREACH(ind_tbl, &priv->ind_tbls, next) {
+ DRV_LOG(DEBUG,
+ "port %u indirection table obj %p still referenced",
+ dev->data->port_id, (void *)ind_tbl);
++ret;
}
return ret;
}
/**
- * Create an indirection table.
+ * Setup an indirection table structure fields.
*
- * @param priv
- * Pointer to private structure.
- * @param queues
- * Queues entering in the indirection table.
- * @param queues_n
- * Number of queues in the array.
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param ind_table
+ * Indirection table to modify.
*
* @return
- * A new indirection table.
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
*/
-struct mlx5_ind_table_ibv*
-mlx5_priv_ind_table_ibv_new(struct priv *priv, uint16_t queues[],
- uint16_t queues_n)
+int
+mlx5_ind_table_obj_setup(struct rte_eth_dev *dev,
+ struct mlx5_ind_table_obj *ind_tbl)
{
- struct mlx5_ind_table_ibv *ind_tbl;
- const unsigned int wq_n = rte_is_power_of_2(queues_n) ?
- log2above(queues_n) :
- priv->ind_table_max_size;
- struct ibv_wq *wq[1 << wq_n];
- unsigned int i;
- unsigned int j;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ uint32_t queues_n = ind_tbl->queues_n;
+ uint16_t *queues = ind_tbl->queues;
+ unsigned int i, j;
+ int ret = 0, err;
+ const unsigned int n = rte_is_power_of_2(queues_n) ?
+ log2above(queues_n) :
+ log2above(priv->config.ind_table_max_size);
- ind_tbl = rte_calloc(__func__, 1, sizeof(*ind_tbl) +
- queues_n * sizeof(uint16_t), 0);
- if (!ind_tbl)
- return NULL;
for (i = 0; i != queues_n; ++i) {
- struct mlx5_rxq_ctrl *rxq =
- mlx5_priv_rxq_get(priv, queues[i]);
-
- if (!rxq)
+ if (!mlx5_rxq_get(dev, queues[i])) {
+ ret = -rte_errno;
goto error;
- wq[i] = rxq->ibv->wq;
- ind_tbl->queues[i] = queues[i];
+ }
}
- ind_tbl->queues_n = queues_n;
- /* Finalise indirection table. */
- for (j = 0; i != (unsigned int)(1 << wq_n); ++i, ++j)
- wq[i] = wq[j];
- ind_tbl->ind_table = ibv_create_rwq_ind_table(
- priv->ctx,
- &(struct ibv_rwq_ind_table_init_attr){
- .log_ind_tbl_size = wq_n,
- .ind_tbl = wq,
- .comp_mask = 0,
- });
- if (!ind_tbl->ind_table)
+ ret = priv->obj_ops.ind_table_new(dev, n, ind_tbl);
+ if (ret)
goto error;
- rte_atomic32_inc(&ind_tbl->refcnt);
- LIST_INSERT_HEAD(&priv->ind_tbls, ind_tbl, next);
- DEBUG("%p: Indirection table %p: refcnt %d", (void *)priv,
- (void *)ind_tbl, rte_atomic32_read(&ind_tbl->refcnt));
- return ind_tbl;
+ __atomic_fetch_add(&ind_tbl->refcnt, 1, __ATOMIC_RELAXED);
+ return 0;
error:
- rte_free(ind_tbl);
- DEBUG("%p cannot create indirection table", (void *)priv);
- return NULL;
+ err = rte_errno;
+ for (j = 0; j < i; j++)
+ mlx5_rxq_release(dev, ind_tbl->queues[j]);
+ rte_errno = err;
+ DEBUG("Port %u cannot setup indirection table.", dev->data->port_id);
+ return ret;
}
/**
- * Get an indirection table.
+ * Create an indirection table.
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
* @param queues
* Queues entering in the indirection table.
* @param queues_n
* Number of queues in the array.
+ * @param standalone
+ * Indirection table for Standalone queue.
*
* @return
- * An indirection table if found.
+ * The Verbs/DevX object initialized, NULL otherwise and rte_errno is set.
*/
-struct mlx5_ind_table_ibv*
-mlx5_priv_ind_table_ibv_get(struct priv *priv, uint16_t queues[],
- uint16_t queues_n)
+static struct mlx5_ind_table_obj *
+mlx5_ind_table_obj_new(struct rte_eth_dev *dev, const uint16_t *queues,
+ uint32_t queues_n, bool standalone)
{
- struct mlx5_ind_table_ibv *ind_tbl;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_ind_table_obj *ind_tbl;
+ int ret;
- LIST_FOREACH(ind_tbl, &priv->ind_tbls, next) {
- if ((ind_tbl->queues_n == queues_n) &&
- (memcmp(ind_tbl->queues, queues,
- ind_tbl->queues_n * sizeof(ind_tbl->queues[0]))
- == 0))
- break;
+ ind_tbl = mlx5_malloc(MLX5_MEM_ZERO, sizeof(*ind_tbl) +
+ queues_n * sizeof(uint16_t), 0, SOCKET_ID_ANY);
+ if (!ind_tbl) {
+ rte_errno = ENOMEM;
+ return NULL;
}
- if (ind_tbl) {
- unsigned int i;
-
- rte_atomic32_inc(&ind_tbl->refcnt);
- DEBUG("%p: Indirection table %p: refcnt %d", (void *)priv,
- (void *)ind_tbl, rte_atomic32_read(&ind_tbl->refcnt));
- for (i = 0; i != ind_tbl->queues_n; ++i)
- mlx5_priv_rxq_get(priv, ind_tbl->queues[i]);
+ ind_tbl->queues_n = queues_n;
+ ind_tbl->queues = (uint16_t *)(ind_tbl + 1);
+ memcpy(ind_tbl->queues, queues, queues_n * sizeof(*queues));
+ ret = mlx5_ind_table_obj_setup(dev, ind_tbl);
+ if (ret < 0) {
+ mlx5_free(ind_tbl);
+ return NULL;
}
+ if (!standalone)
+ LIST_INSERT_HEAD(&priv->ind_tbls, ind_tbl, next);
return ind_tbl;
}
/**
- * Release an indirection table.
+ * Modify an indirection table.
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
* @param ind_table
- * Indirection table to release.
+ * Indirection table to modify.
+ * @param queues
+ * Queues replacement for the indirection table.
+ * @param queues_n
+ * Number of queues in the array.
+ * @param standalone
+ * Indirection table for Standalone queue.
*
* @return
- * 0 on success, errno value on failure.
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
*/
int
-mlx5_priv_ind_table_ibv_release(struct priv *priv,
- struct mlx5_ind_table_ibv *ind_tbl)
+mlx5_ind_table_obj_modify(struct rte_eth_dev *dev,
+ struct mlx5_ind_table_obj *ind_tbl,
+ uint16_t *queues, const uint32_t queues_n,
+ bool standalone)
{
- unsigned int i;
-
- DEBUG("%p: Indirection table %p: refcnt %d", (void *)priv,
- (void *)ind_tbl, rte_atomic32_read(&ind_tbl->refcnt));
- if (rte_atomic32_dec_and_test(&ind_tbl->refcnt))
- claim_zero(ibv_destroy_rwq_ind_table(ind_tbl->ind_table));
- for (i = 0; i != ind_tbl->queues_n; ++i)
- claim_nonzero(mlx5_priv_rxq_release(priv, ind_tbl->queues[i]));
- if (!rte_atomic32_read(&ind_tbl->refcnt)) {
- LIST_REMOVE(ind_tbl, next);
- rte_free(ind_tbl);
- return 0;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ unsigned int i, j;
+ int ret = 0, err;
+ const unsigned int n = rte_is_power_of_2(queues_n) ?
+ log2above(queues_n) :
+ log2above(priv->config.ind_table_max_size);
+
+ MLX5_ASSERT(standalone);
+ RTE_SET_USED(standalone);
+ if (__atomic_load_n(&ind_tbl->refcnt, __ATOMIC_RELAXED) > 1) {
+ /*
+ * Modification of indirection ntables having more than 1
+ * reference unsupported. Intended for standalone indirection
+ * tables only.
+ */
+ DEBUG("Port %u cannot modify indirection table (refcnt> 1).",
+ dev->data->port_id);
+ rte_errno = EINVAL;
+ return -rte_errno;
+ }
+ for (i = 0; i != queues_n; ++i) {
+ if (!mlx5_rxq_get(dev, queues[i])) {
+ ret = -rte_errno;
+ goto error;
+ }
}
- return EBUSY;
+ MLX5_ASSERT(priv->obj_ops.ind_table_modify);
+ ret = priv->obj_ops.ind_table_modify(dev, n, queues, queues_n, ind_tbl);
+ if (ret)
+ goto error;
+ for (j = 0; j < ind_tbl->queues_n; j++)
+ mlx5_rxq_release(dev, ind_tbl->queues[j]);
+ ind_tbl->queues_n = queues_n;
+ ind_tbl->queues = queues;
+ return 0;
+error:
+ err = rte_errno;
+ for (j = 0; j < i; j++)
+ mlx5_rxq_release(dev, ind_tbl->queues[j]);
+ rte_errno = err;
+ DEBUG("Port %u cannot setup indirection table.", dev->data->port_id);
+ return ret;
}
/**
- * Verify the Rx Queue list is empty
+ * Match an Rx Hash queue.
*
- * @param priv
- * Pointer to private structure.
+ * @param list
+ * Cache list pointer.
+ * @param entry
+ * Hash queue entry pointer.
+ * @param cb_ctx
+ * Context of the callback function.
*
- * @return the number of object not released.
+ * @return
+ * 0 if match, none zero if not match.
*/
int
-mlx5_priv_ind_table_ibv_verify(struct priv *priv)
+mlx5_hrxq_match_cb(struct mlx5_cache_list *list,
+ struct mlx5_cache_entry *entry,
+ void *cb_ctx)
{
- struct mlx5_ind_table_ibv *ind_tbl;
- int ret = 0;
-
- LIST_FOREACH(ind_tbl, &priv->ind_tbls, next) {
- DEBUG("%p: Verbs indirection table %p still referenced",
- (void *)priv, (void *)ind_tbl);
- ++ret;
- }
- return ret;
+ struct rte_eth_dev *dev = list->ctx;
+ struct mlx5_flow_cb_ctx *ctx = cb_ctx;
+ struct mlx5_flow_rss_desc *rss_desc = ctx->data;
+ struct mlx5_hrxq *hrxq = container_of(entry, typeof(*hrxq), entry);
+ struct mlx5_ind_table_obj *ind_tbl;
+
+ if (hrxq->rss_key_len != rss_desc->key_len ||
+ memcmp(hrxq->rss_key, rss_desc->key, rss_desc->key_len) ||
+ hrxq->hash_fields != rss_desc->hash_fields)
+ return 1;
+ ind_tbl = mlx5_ind_table_obj_get(dev, rss_desc->queue,
+ rss_desc->queue_num);
+ if (ind_tbl)
+ mlx5_ind_table_obj_release(dev, ind_tbl, hrxq->standalone);
+ return ind_tbl != hrxq->ind_table;
}
/**
- * Create an Rx Hash queue.
+ * Modify an Rx Hash queue configuration.
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param hrxq
+ * Index to Hash Rx queue to modify.
* @param rss_key
* RSS key for the Rx hash queue.
* @param rss_key_len
* @param hash_fields
* Verbs protocol hash field to make the RSS on.
* @param queues
- * Queues entering in hash queue.
+ * Queues entering in hash queue. In case of empty hash_fields only the
+ * first queue index will be taken for the indirection table.
* @param queues_n
* Number of queues.
*
* @return
- * An hash Rx queue on success.
+ * 0 on success, a negative errno value otherwise and rte_errno is set.
*/
-struct mlx5_hrxq*
-mlx5_priv_hrxq_new(struct priv *priv, uint8_t *rss_key, uint8_t rss_key_len,
- uint64_t hash_fields, uint16_t queues[], uint16_t queues_n)
+int
+mlx5_hrxq_modify(struct rte_eth_dev *dev, uint32_t hrxq_idx,
+ const uint8_t *rss_key, uint32_t rss_key_len,
+ uint64_t hash_fields,
+ const uint16_t *queues, uint32_t queues_n)
{
- struct mlx5_hrxq *hrxq;
- struct mlx5_ind_table_ibv *ind_tbl;
- struct ibv_qp *qp;
+ int err;
+ struct mlx5_ind_table_obj *ind_tbl = NULL;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_hrxq *hrxq =
+ mlx5_ipool_get(priv->sh->ipool[MLX5_IPOOL_HRXQ], hrxq_idx);
+ int ret;
+
+ if (!hrxq) {
+ rte_errno = EINVAL;
+ return -rte_errno;
+ }
+ /* validations */
+ if (hrxq->rss_key_len != rss_key_len) {
+ /* rss_key_len is fixed size 40 byte & not supposed to change */
+ rte_errno = EINVAL;
+ return -rte_errno;
+ }
+ queues_n = hash_fields ? queues_n : 1;
+ if (mlx5_ind_table_obj_match_queues(hrxq->ind_table,
+ queues, queues_n)) {
+ ind_tbl = hrxq->ind_table;
+ } else {
+ if (hrxq->standalone) {
+ /*
+ * Replacement of indirection table unsupported for
+ * stanalone hrxq objects (used by shared RSS).
+ */
+ rte_errno = ENOTSUP;
+ return -rte_errno;
+ }
+ ind_tbl = mlx5_ind_table_obj_get(dev, queues, queues_n);
+ if (!ind_tbl)
+ ind_tbl = mlx5_ind_table_obj_new(dev, queues, queues_n,
+ hrxq->standalone);
+ }
+ if (!ind_tbl) {
+ rte_errno = ENOMEM;
+ return -rte_errno;
+ }
+ MLX5_ASSERT(priv->obj_ops.hrxq_modify);
+ ret = priv->obj_ops.hrxq_modify(dev, hrxq, rss_key,
+ hash_fields, ind_tbl);
+ if (ret) {
+ rte_errno = errno;
+ goto error;
+ }
+ if (ind_tbl != hrxq->ind_table) {
+ MLX5_ASSERT(!hrxq->standalone);
+ mlx5_ind_table_obj_release(dev, hrxq->ind_table,
+ hrxq->standalone);
+ hrxq->ind_table = ind_tbl;
+ }
+ hrxq->hash_fields = hash_fields;
+ memcpy(hrxq->rss_key, rss_key, rss_key_len);
+ return 0;
+error:
+ err = rte_errno;
+ if (ind_tbl != hrxq->ind_table) {
+ MLX5_ASSERT(!hrxq->standalone);
+ mlx5_ind_table_obj_release(dev, ind_tbl, hrxq->standalone);
+ }
+ rte_errno = err;
+ return -rte_errno;
+}
+
+static void
+__mlx5_hrxq_remove(struct rte_eth_dev *dev, struct mlx5_hrxq *hrxq)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+
+#ifdef HAVE_IBV_FLOW_DV_SUPPORT
+ mlx5_glue->destroy_flow_action(hrxq->action);
+#endif
+ priv->obj_ops.hrxq_destroy(hrxq);
+ if (!hrxq->standalone) {
+ mlx5_ind_table_obj_release(dev, hrxq->ind_table,
+ hrxq->standalone);
+ }
+ mlx5_ipool_free(priv->sh->ipool[MLX5_IPOOL_HRXQ], hrxq->idx);
+}
+
+/**
+ * Release the hash Rx queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param hrxq
+ * Index to Hash Rx queue to release.
+ *
+ * @param list
+ * Cache list pointer.
+ * @param entry
+ * Hash queue entry pointer.
+ */
+void
+mlx5_hrxq_remove_cb(struct mlx5_cache_list *list,
+ struct mlx5_cache_entry *entry)
+{
+ struct rte_eth_dev *dev = list->ctx;
+ struct mlx5_hrxq *hrxq = container_of(entry, typeof(*hrxq), entry);
+
+ __mlx5_hrxq_remove(dev, hrxq);
+}
- ind_tbl = mlx5_priv_ind_table_ibv_get(priv, queues, queues_n);
+static struct mlx5_hrxq *
+__mlx5_hrxq_create(struct rte_eth_dev *dev,
+ struct mlx5_flow_rss_desc *rss_desc)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ const uint8_t *rss_key = rss_desc->key;
+ uint32_t rss_key_len = rss_desc->key_len;
+ bool standalone = !!rss_desc->shared_rss;
+ const uint16_t *queues =
+ standalone ? rss_desc->const_q : rss_desc->queue;
+ uint32_t queues_n = rss_desc->queue_num;
+ struct mlx5_hrxq *hrxq = NULL;
+ uint32_t hrxq_idx = 0;
+ struct mlx5_ind_table_obj *ind_tbl = rss_desc->ind_tbl;
+ int ret;
+
+ queues_n = rss_desc->hash_fields ? queues_n : 1;
+ if (!ind_tbl)
+ ind_tbl = mlx5_ind_table_obj_get(dev, queues, queues_n);
if (!ind_tbl)
- ind_tbl = mlx5_priv_ind_table_ibv_new(priv, queues, queues_n);
+ ind_tbl = mlx5_ind_table_obj_new(dev, queues, queues_n,
+ standalone);
if (!ind_tbl)
return NULL;
- qp = ibv_create_qp_ex(
- priv->ctx,
- &(struct ibv_qp_init_attr_ex){
- .qp_type = IBV_QPT_RAW_PACKET,
- .comp_mask =
- IBV_QP_INIT_ATTR_PD |
- IBV_QP_INIT_ATTR_IND_TABLE |
- IBV_QP_INIT_ATTR_RX_HASH,
- .rx_hash_conf = (struct ibv_rx_hash_conf){
- .rx_hash_function = IBV_RX_HASH_FUNC_TOEPLITZ,
- .rx_hash_key_len = rss_key_len,
- .rx_hash_key = rss_key,
- .rx_hash_fields_mask = hash_fields,
- },
- .rwq_ind_tbl = ind_tbl->ind_table,
- .pd = priv->pd,
- });
- if (!qp)
- goto error;
- hrxq = rte_calloc(__func__, 1, sizeof(*hrxq) + rss_key_len, 0);
+ hrxq = mlx5_ipool_zmalloc(priv->sh->ipool[MLX5_IPOOL_HRXQ], &hrxq_idx);
if (!hrxq)
goto error;
+ hrxq->standalone = standalone;
+ hrxq->idx = hrxq_idx;
hrxq->ind_table = ind_tbl;
- hrxq->qp = qp;
hrxq->rss_key_len = rss_key_len;
- hrxq->hash_fields = hash_fields;
+ hrxq->hash_fields = rss_desc->hash_fields;
memcpy(hrxq->rss_key, rss_key, rss_key_len);
- rte_atomic32_inc(&hrxq->refcnt);
- LIST_INSERT_HEAD(&priv->hrxqs, hrxq, next);
- DEBUG("%p: Hash Rx queue %p: refcnt %d", (void *)priv,
- (void *)hrxq, rte_atomic32_read(&hrxq->refcnt));
+ ret = priv->obj_ops.hrxq_new(dev, hrxq, rss_desc->tunnel);
+ if (ret < 0)
+ goto error;
return hrxq;
error:
- mlx5_priv_ind_table_ibv_release(priv, ind_tbl);
- if (qp)
- claim_zero(ibv_destroy_qp(qp));
+ if (!rss_desc->ind_tbl)
+ mlx5_ind_table_obj_release(dev, ind_tbl, standalone);
+ if (hrxq)
+ mlx5_ipool_free(priv->sh->ipool[MLX5_IPOOL_HRXQ], hrxq_idx);
return NULL;
}
+/**
+ * Create an Rx Hash queue.
+ *
+ * @param list
+ * Cache list pointer.
+ * @param entry
+ * Hash queue entry pointer.
+ * @param cb_ctx
+ * Context of the callback function.
+ *
+ * @return
+ * queue entry on success, NULL otherwise.
+ */
+struct mlx5_cache_entry *
+mlx5_hrxq_create_cb(struct mlx5_cache_list *list,
+ struct mlx5_cache_entry *entry __rte_unused,
+ void *cb_ctx)
+{
+ struct rte_eth_dev *dev = list->ctx;
+ struct mlx5_flow_cb_ctx *ctx = cb_ctx;
+ struct mlx5_flow_rss_desc *rss_desc = ctx->data;
+ struct mlx5_hrxq *hrxq;
+
+ hrxq = __mlx5_hrxq_create(dev, rss_desc);
+ return hrxq ? &hrxq->entry : NULL;
+}
+
/**
* Get an Rx Hash queue.
*
- * @param priv
- * Pointer to private structure.
- * @param rss_conf
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param rss_desc
* RSS configuration for the Rx hash queue.
- * @param queues
- * Queues entering in hash queue.
- * @param queues_n
- * Number of queues.
*
* @return
- * An hash Rx queue on success.
+ * An hash Rx queue index on success.
*/
-struct mlx5_hrxq*
-mlx5_priv_hrxq_get(struct priv *priv, uint8_t *rss_key, uint8_t rss_key_len,
- uint64_t hash_fields, uint16_t queues[], uint16_t queues_n)
+uint32_t mlx5_hrxq_get(struct rte_eth_dev *dev,
+ struct mlx5_flow_rss_desc *rss_desc)
{
+ struct mlx5_priv *priv = dev->data->dev_private;
struct mlx5_hrxq *hrxq;
+ struct mlx5_cache_entry *entry;
+ struct mlx5_flow_cb_ctx ctx = {
+ .data = rss_desc,
+ };
- LIST_FOREACH(hrxq, &priv->hrxqs, next) {
- struct mlx5_ind_table_ibv *ind_tbl;
-
- if (hrxq->rss_key_len != rss_key_len)
- continue;
- if (memcmp(hrxq->rss_key, rss_key, rss_key_len))
- continue;
- if (hrxq->hash_fields != hash_fields)
- continue;
- ind_tbl = mlx5_priv_ind_table_ibv_get(priv, queues, queues_n);
- if (!ind_tbl)
- continue;
- if (ind_tbl != hrxq->ind_table) {
- mlx5_priv_ind_table_ibv_release(priv, ind_tbl);
- continue;
- }
- rte_atomic32_inc(&hrxq->refcnt);
- DEBUG("%p: Hash Rx queue %p: refcnt %d", (void *)priv,
- (void *)hrxq, rte_atomic32_read(&hrxq->refcnt));
- return hrxq;
+ if (rss_desc->shared_rss) {
+ hrxq = __mlx5_hrxq_create(dev, rss_desc);
+ } else {
+ entry = mlx5_cache_register(&priv->hrxqs, &ctx);
+ if (!entry)
+ return 0;
+ hrxq = container_of(entry, typeof(*hrxq), entry);
}
- return NULL;
+ if (hrxq)
+ return hrxq->idx;
+ return 0;
}
/**
* Release the hash Rx queue.
*
- * @param priv
- * Pointer to private structure.
- * @param hrxq
- * Pointer to Hash Rx queue to release.
+ * @param dev
+ * Pointer to Ethernet device.
+ * @param hrxq_idx
+ * Index to Hash Rx queue to release.
*
* @return
- * 0 on success, errno value on failure.
+ * 1 while a reference on it exists, 0 when freed.
*/
-int
-mlx5_priv_hrxq_release(struct priv *priv, struct mlx5_hrxq *hrxq)
+int mlx5_hrxq_release(struct rte_eth_dev *dev, uint32_t hrxq_idx)
{
- DEBUG("%p: Hash Rx queue %p: refcnt %d", (void *)priv,
- (void *)hrxq, rte_atomic32_read(&hrxq->refcnt));
- if (rte_atomic32_dec_and_test(&hrxq->refcnt)) {
- claim_zero(ibv_destroy_qp(hrxq->qp));
- mlx5_priv_ind_table_ibv_release(priv, hrxq->ind_table);
- LIST_REMOVE(hrxq, next);
- rte_free(hrxq);
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_hrxq *hrxq;
+
+ hrxq = mlx5_ipool_get(priv->sh->ipool[MLX5_IPOOL_HRXQ], hrxq_idx);
+ if (!hrxq)
return 0;
+ if (!hrxq->standalone)
+ return mlx5_cache_unregister(&priv->hrxqs, &hrxq->entry);
+ __mlx5_hrxq_remove(dev, hrxq);
+ return 0;
+}
+
+/**
+ * Create a drop Rx Hash queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * The Verbs/DevX object initialized, NULL otherwise and rte_errno is set.
+ */
+struct mlx5_hrxq *
+mlx5_drop_action_create(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_hrxq *hrxq = NULL;
+ int ret;
+
+ if (priv->drop_queue.hrxq)
+ return priv->drop_queue.hrxq;
+ hrxq = mlx5_malloc(MLX5_MEM_ZERO, sizeof(*hrxq), 0, SOCKET_ID_ANY);
+ if (!hrxq) {
+ DRV_LOG(WARNING,
+ "Port %u cannot allocate memory for drop queue.",
+ dev->data->port_id);
+ rte_errno = ENOMEM;
+ goto error;
+ }
+ priv->drop_queue.hrxq = hrxq;
+ hrxq->ind_table = mlx5_malloc(MLX5_MEM_ZERO, sizeof(*hrxq->ind_table),
+ 0, SOCKET_ID_ANY);
+ if (!hrxq->ind_table) {
+ rte_errno = ENOMEM;
+ goto error;
}
- claim_nonzero(mlx5_priv_ind_table_ibv_release(priv, hrxq->ind_table));
- return EBUSY;
+ ret = priv->obj_ops.drop_action_create(dev);
+ if (ret < 0)
+ goto error;
+ return hrxq;
+error:
+ if (hrxq) {
+ if (hrxq->ind_table)
+ mlx5_free(hrxq->ind_table);
+ priv->drop_queue.hrxq = NULL;
+ mlx5_free(hrxq);
+ }
+ return NULL;
+}
+
+/**
+ * Release a drop hash Rx queue.
+ *
+ * @param dev
+ * Pointer to Ethernet device.
+ */
+void
+mlx5_drop_action_destroy(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_hrxq *hrxq = priv->drop_queue.hrxq;
+
+ if (!priv->drop_queue.hrxq)
+ return;
+ priv->obj_ops.drop_action_destroy(dev);
+ mlx5_free(priv->drop_queue.rxq);
+ mlx5_free(hrxq->ind_table);
+ mlx5_free(hrxq);
+ priv->drop_queue.rxq = NULL;
+ priv->drop_queue.hrxq = NULL;
}
/**
* Verify the Rx Queue list is empty
*
- * @param priv
- * Pointer to private structure.
+ * @param dev
+ * Pointer to Ethernet device.
+ *
+ * @return
+ * The number of object not released.
+ */
+uint32_t
+mlx5_hrxq_verify(struct rte_eth_dev *dev)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+
+ return mlx5_cache_list_get_entry_num(&priv->hrxqs);
+}
+
+/**
+ * Set the Rx queue timestamp conversion parameters
*
- * @return the number of object not released.
+ * @param[in] dev
+ * Pointer to the Ethernet device structure.
*/
-int
-mlx5_priv_hrxq_ibv_verify(struct priv *priv)
+void
+mlx5_rxq_timestamp_set(struct rte_eth_dev *dev)
{
- struct mlx5_hrxq *hrxq;
- int ret = 0;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_dev_ctx_shared *sh = priv->sh;
+ struct mlx5_rxq_data *data;
+ unsigned int i;
- LIST_FOREACH(hrxq, &priv->hrxqs, next) {
- DEBUG("%p: Verbs Hash Rx queue %p still referenced",
- (void *)priv, (void *)hrxq);
- ++ret;
+ for (i = 0; i != priv->rxqs_n; ++i) {
+ if (!(*priv->rxqs)[i])
+ continue;
+ data = (*priv->rxqs)[i];
+ data->sh = sh;
+ data->rt_timestamp = priv->config.rt_timestamp;
}
- return ret;
}